Differences

This shows you the differences between two versions of the page.

--- provenance:privacy:fingerprinting [2026/08/12 10:07] – Add section 10: the Fable review pass, the windowed-guard defect it exposed (29 corpus figures stale outside the audited heading window across six pages), the hardened check_page_numbers.mjs, and the page-specific fixes. Authored by Claude. karel.kubicek.claude
+++ provenance:privacy:fingerprinting [2026/09/21 13:30] (current) – Quote-check refresh 2026-09-21: re-ran quote_check.mjs --detection fingerprint with the pypdf fallback; 92 below threshold -> 50 rescued + 42 below in both; new provenance section. Authored by Claude karel.kubicek.claude
@@ Line 11: / Line 11: @@
 ^ Item ^ Value ^
 | Content page | [[privacy:fingerprinting]] |
-| Report script | ''scripts/report_fingerprinting.mjs'' (''--wiki'' emits DokuWiki tables) |
+| Report script | ''scripts/report_fingerprinting.mjs'' (''%%--wiki%%'' emits DokuWiki tables) |
 | Fold | ''scripts/fp_fold.mjs'' — three passes: STRONG patterns, GENERIC resolved by paper context, HAND table |
-| Quote verification | ''scripts/quote_check.mjs --detection fingerprint'' |
+| Quote verification | ''%%scripts/quote_check.mjs --detection fingerprint%%'' |
 | Data | ''data/extract/run1/extractions.jsonl'', 5,859 papers, 7 venues, 2010–2026 |
-| Refreshed | 2026-08-12 |
+| Refreshed | 2026-08-12 (corpus), 2026-09-03 (review pass, see §11) |
 ===== 2. Populations and denominators =====
@@ Line 138: / Line 138: @@
 ===== 6. Quotes checked =====
-//Recorded, 2026-08-12.//
+//Recorded, 2026-08-12. Re-run 2026-09-21 with the PDF fallback — see //Quote-check refresh, 2026-09-21// at the foot of this page for what moved and why.//
 <code>
 $ node scripts/quote_check.mjs --detection 'fingerprint'
-quotes checked: 312 exact, 140 partial (>=60% of 5-word windows),
+quotes checked: 312 exact, 140 partial (>=60% of 5-word windows), 50 rescued from the PDF, 42 below threshold in both renderings, 0 with no full text on disk.
- below threshold, 0 with no full text on disk.
 </code>
-**92 below threshold were not all read.** Five were sampled and checked by hand against ''paper.cols.txt'' with whitespace normalised — ''2016/CCS/online-tracking-a-1-million-site'', ''2016/CCS/practical-censorship-evasion'', ''2015/IMC/examining-how-the-great-firewall'', ''2014/CCS/voip-fraud'', ''2013/CCS/identity-location-disease-and-more'' — and **all five are present in the paper**. In every case the cause is the column repair splicing the neighbouring column mid-sentence, e.g.:
+**The old figure was 92 below threshold; 50 of those 92 are defects in the stored text, not in the extraction.** They are below threshold against the rendering the extractor read and at or above it against an independent ''pypdf'' rendering of the same ''paper.pdf''. ''exact'' (312) and ''partial'' (140) did not move. What has to be read by hand is the **42** that are below threshold in **both** renderings.
+**Not all of the 42 were read.** Before the fallback existed, five of the then-92 were sampled and checked by hand against ''paper.cols.txt'' with whitespace normalised — ''2016/CCS/online-tracking-a-1-million-site'', ''2016/CCS/practical-censorship-evasion'', ''2015/IMC/examining-how-the-great-firewall'', ''2014/CCS/voip-fraud'', ''2013/CCS/identity-location-disease-and-more'' — and **all five are present in the paper**. The fallback now reaches three of those five on its own (''practical-censorship-evasion'', ''examining-how-the-great-firewall'', ''voip-fraud''); ''online-tracking-a-1-million-site'' and ''identity-location-disease-and-more'' are still below threshold in both renderings and are in the 42, so the hand-read result is the only evidence for them. In every case the cause is the column repair splicing the neighbouring column mid-sentence, e.g.:
 <code>
@@ Line 167: / Line 168: @@
   * The "2025 policy reversal" section rests on vendor and standards announcements checked by the original run in August 2026. **It is the part of this page most likely to be stale first**, and nothing in this refresh touched it.
-**Rejected:** not recorded for the original run.
+**Rejected:** not recorded for the original run — a gap, and the reason §11 records them explicitly.
 ===== 8. What could not be established =====
@@ Line 178: / Line 179: @@
   * **Whether the 103 technique-only exclusions are all correctly excluded.** The rule is defensible and it has never been audited paper by paper.
-===== 10. Review pass, 2026-08-12 =====
+===== 9. Review pass, 2026-08-12 =====
 //Recorded.// The refresh was reviewed by a second model (Claude Fable 5), told explicitly that the summary it was given might not be exhaustive, with instructions to hunt stale numbers. It found a **systematic defect, not a scatter of typos**, and it is worth stating because it will recur on the next refresh:
@@ Line 185: / Line 186: @@
   * The matcher was **substring**, not word-boundary, so ''report.includes('59')'' was satisfied by ''11.59 bits''. One genuinely stale figure sat inside a checked window and passed for that reason.
-Both are fixed in ''scripts/check_page_numbers.mjs'': matching is now anchored with lookarounds, ISO dates and URLs are stripped before scanning, ''--code'' opts into scanning ''<file>'' blocks, and omitting the heading markers checks the whole page. **Run it windowed //and// whole-page.** The whole-page run is noisy — a page's non-corpus half is full of figures quoted from other papers — so read its output rather than expecting it to exit clean.
+Both are fixed in ''scripts/check_page_numbers.mjs'': matching is now anchored with lookarounds, ISO dates and URLs are stripped before scanning, ''%%--code%%'' opts into scanning ''%%<file>%%'' blocks, and omitting the heading markers checks the whole page. **Run it windowed //and// whole-page.** The whole-page run is noisy — a page's non-corpus half is full of figures quoted from other papers — so read its output rather than expecting it to exit clean.
-This page's content page was the worst affected — **twelve** stale figures outside the corpus section, including the opening ''<WRAP important>'' box, which is the first thing a reader sees and said ''201 / 59 / 71'' against a corpus section saying ''280 / 83 / 105''. Also fixed: OpenWPM **12 → 15**, Tracker Radar Collector **3 → 6** (it contradicted the page's own tool table), FingerprintJS **9 → 14**, ''puppeteer-extra-plugin-stealth'' **1 → 2**, legal **7 of 59 → 10 of 83**, and two Related-Pages shares.
+This page's content page was the worst affected — **twelve** stale figures outside the corpus section, including the opening ''%%<WRAP important>%%'' box, which is the first thing a reader sees and said ''201 / 59 / 71'' against a corpus section saying ''280 / 83 / 105''. Also fixed: OpenWPM **12 → 15**, Tracker Radar Collector **3 → 6** (it contradicted the page's own tool table), FingerprintJS **9 → 14**, ''puppeteer-extra-plugin-stealth'' **1 → 2**, legal **7 of 59 → 10 of 83**, and two Related-Pages shares.
 **Two claims were not stale but false, and had to be re-derived rather than renumbered.** Both are recorded here because they go beyond the mechanical brief:
@@ Line 195: / Line 196: @@
 Both papers were added to [[literature:bibliography]] as ''luo2025_canvassing'' and ''annamalai2025_beyond'' (duplicate-key check first, cache purged, keys confirmed rendering).
-===== 11. Run log =====
+===== 10. Run log =====
 ^ ^ ^
@@ Line 206: / Line 207: @@
 | Mistake caught in review | The first edit to ''fp_fold.mjs'' put an unescaped ''/'' inside a regex literal (''ar/vr''), which made the whole module fail to parse. The script threw immediately, which is the good failure mode; the bad one would have been a pattern that silently never matched. |
 | Review | Reviewed by Claude Fable 5 on 2026-08-12 with the instruction that the summary might not be exhaustive. It found the windowed-guard defect in §10 and 12 stale figures on this page plus two claims that had become false. All fixes were applied and re-saved the same day. |
+===== 11. Review pass, 2026-09-03 — the generic (Fable) reviewer =====
+//Recorded.// The step-9 generic review was **skipped for budget** when this page was written, and was run on 2026-09-03 as its own task. The reviewer (Claude Fable) was given the content page, the provenance page, ''report_fingerprinting.mjs'', ''fp_fold.mjs'' and a re-run of the report against the current corpus, and was told explicitly that the author's context might not be exhaustive and that the three focused passes (figures-vs-script, citations-and-quotes, external currency) had already run — so its budget should go on what none of them was looking for. No checklist was given.
+The corpus did **not** change between 2026-08-12 and this pass: still ''data/extract/run1'', 5,859 papers. Every figure below was re-derived, not carried over.
+==== 11.1 Found before the reviewer returned ====
+Three defects were found by the author while preparing the reviewer's bundle. They are recorded here because two of them are the kind a reviewer brief does not own.
+^ # ^ Defect ^ Evidence ^ Action ^
+| A | ''%%--wiki%%'', ''%%--detection%%'' and ''%%--code%%'' were written as ''%%--flag%%'' inside ''%%''…''%%'' monospace. DokuWiki's typography converts the double hyphen to an **en-dash**, so five published command-line flags across these two pages rendered as ''%%–wiki%%'' etc. and fail on copy-paste. | Rendered DOM: ''%%<code>–wiki</code>%%''. The source regex alone does not show this — the damage only exists after rendering. | Fixed by wrapping each literal in ''%%''%%…%%''%%'', the idiom already used on ''provenance:artifacts'' and ''security:dns''. **Site-wide**: the same defect is live on at least ''programming:docker'', ''programming:traffic_files'' and ''provenance:artifacts'', and 72 local page drafts match the source pattern. Filed as its own task item rather than fixed here. |
+| B | The ''%%<file javascript fp_fold.mjs>%%'' block on the content page was an **abridged rewrite** of the module, not the module: 52 lines against 184, with three families (''media-content'', ''device-hardware'', ''evasion-environment'') deleted and the ''traffic-analysis'' pattern cut from ~40 alternatives to 7 — under a ''%%<file>%%'' block named ''fp_fold.mjs'', i.e. downloadable under the real filename. | Reconstructed the published excerpt into a runnable module and ran it against the same 280 papers. It reproduces **none** of the ten rows of the page's own family table. | Fixed: the block is now the committed file **verbatim** (diffed byte-for-byte before saving). |
+| C | ''19.6%'' as the corpus-wide statefulness rate in the //Defences// section, against ''21.9%'' in the page's own crawl-config table three sections earlier. | ''report_fingerprinting.mjs'' line 143: ''statefulness 21 36.8% 21.9%''. | Fixed. Independently found by the reviewer as its finding 8. |
+**What B would have produced**, run against the same 280 papers (left: the committed module and the published table; right: the code the page actually published):
+<code>
+family                   real  published-excerpt
+traffic-analysis          105         89
+browser-device             83         65
+network-stack              38         23
+physical-forensic          28          9
+biometric                  16          2
+device-hardware            13          3
+media-content               8          0
+unmapped                    6         92
+evasion-environment         2          0
+llm-model                   1          1
+</code>
+The unmapped row is the one that matters: the page's headline methodological claim is that the fold leaves **7 tuples in 6 papers** unassigned. Under the published code it leaves **92 papers**, 32.9% of the population. The page was honest about the block being an excerpt — it said so on its line 15 — but an excerpt that changes every number beneath it is not an audit trail, and "the full module … is in the wiki repository" pointed at a repository neither page links because none is public.
+==== 11.2 The reviewer's findings, and what was done with each ====
+Fifteen findings. **Thirteen accepted, two accepted in part.** Every load-bearing claim was verified against a primary source before acting — the reviewer was right on each one that was checked, which is worth recording as much as any rejection would be.
+^ # ^ Severity ^ Finding ^ Verified how ^ Action ^
+| 1 | **blocking** | The open question "none in the corpus tests whether a CMP's consent signal changes fingerprinting behaviour on the same site" is **refuted by a paper cited on the page**. | Read ''2021/WWW/user-tracking-in-the-post-cookie-era…/paper.cols.txt''. §4.3 verbatim: //"we set out to explore whether websites track users differently via browser fingerprinting, given the different user responses to the requested cookie consent"//. Table 6 verbatim: //"No Action 279 1.03% Reject All 285 1.05% Accept All 330 1.21% … In all 3 cases 247"//. Denominator located in the same paper rather than back-calculated: //"Websites with a CMP and no error in all three consent actions 27,180"//; 279/27,180 = 1.03% checks out. | **Accepted.** Question rewritten to state the published result with its denominator and narrow what is actually open (the post-2025 repeat, and whether any refusal signal is honoured). |
+| 2 | should-fix | Three places disagree about who measured the automated-crawl gap: FP-Fed "explicitly measures" it, yet Annamalai 2025 is credited with "the first published size". | Read ''2024/NDSS/fp-fed…/paper.cols.txt'': //"we also conduct a small-scale study on the top 300 domains from the Tranco ranking list and find that crawls involving real users (logging in, solving CAPTCHAs, etc.) can capture 3 times more fingerprinting scripts than automated/centralized ones"//. So FP-Fed did publish a size. | **Accepted.** All three passages rewritten: FP-Fed is the top-300 pilot at 3×, Annamalai 2025 is the first size //at scale//, and both are noted to vary interaction, login and consent together — so neither isolates headless-vs-headful. |
+| 3 | should-fix | The page recommends OpenWPM on a corpus rank while warning elsewhere that the browser is a defence configuration, and never connects the two. | ''programming:crawler:openwpm'' on this wiki, which reads the repository directly: unbranded **Firefox only**; each release pins one build (v0.35.0, 2026-06-17 → Firefox 152); and ''js_instrument=True'' means the fixed ''collection_fingerprinting'' list of 14 objects, **content-unchanged since 2020-07-08**, with no ''WebGLRenderingContext'', no ''document.fonts'', and ''window.screen'' only for ''colorDepth''/''pixelDepth''. | **Accepted, and strengthened.** The reviewer's suggested fix was about fingerprinting-protection prefs; the sharper and verifiable defect is the instrumentation gap — OpenWPM's default JS collection misses three of the surfaces in this page's own table. The tooling section now opens "pick by question, not by citation count" and says so. |
+| 4 | should-fix | The tool fold has a silent ''if (!fam) continue;'' and prints no residue, in violation of the house rule; the drop hides the crawl drivers and VisibleV8. | Ran the fold's drop set over the 83 papers: 398 distinct names dropped, 55 in ≥2 papers — Selenium 13, Puppeteer 12, CDP 5, VisibleV8 5, Docker 5, mitmproxy 4, Playwright 3. VisibleV8 at 5 outranks FP-Inspector (4) and AmIUnique (4) in the published table. | **Accepted.** ''report_fingerprinting.mjs'' now collects and prints the residue in full (see the output below). The page gained a ''%%<WRAP info>%%'' naming the drivers, and VisibleV8 is now a tooling entry in its own right. |
+| 5 | should-fix | Two overstated bullets: "a keyword search … is about 30% precise" (that is the precision of the //extraction//, not of a keyword search) and "both literatures are the same size and both are growing" (105 vs 83, and browser FP goes 16.0 → 14.8 → 15.2, i.e. flat). | The second half is checkable on the page's own numbers. | **Accepted**, both rewritten. Browser fingerprinting is now described as flat near 15 per 1,000 since 2018, which is what the series says. |
+| 6 | should-fix | "They do not cite each other" is unsupported — the extraction has no citation graph — and "reviewers from one side routinely read submissions from the other" is anecdote. | The schema has no citation field. | **Accepted.** The methods contrast is kept; the citation claim is replaced with an explicit statement that this page cannot answer it. |
+| 7 | should-fix | "Filter lists were not built to label fingerprinting" is **wrong for Disconnect**; and the page never tells the reader which datasets exist. | Fetched ''disconnect-blacklist.json'' from ''mozilla-services/shavar-prod-lists'' on 2026-09-03: 11 categories including ''FingerprintingInvasive'' (281 entries) and ''FingerprintingGeneral'' (96). Fetched ''docs/DATA_MODEL.md'' and ''README.md'' from ''duckduckgo/tracker-radar'': per-domain and per-resource ''fingerprinting'' score and an ''apis'' list of //"browser APIs accessed by this resource which are commonly used in fingerprinting"//; CC BY-NC-SA 4.0. Sampled scores: doubleclick.net 3, cloudflare.com 2, facebook.net 2, criteo.com 1. | **Accepted.** The filter-list paragraph now separates EasyList/EasyPrivacy (no fingerprinting category) from Disconnect (two of them, and papers should say which they used), and a new //Datasets you can start from// subsection was added. |
+| 8 | should-fix | 19.6% vs 21.9%. | See §11.1 C. | **Accepted**; had already been fixed. |
+| 9 | should-fix | The fold subsection, code block and output sit between the reader and //Use in Publications//; the published excerpt does not reproduce the output beneath it; no repository is linked; the 83-paper list is promised twice and never published; and this provenance page promises "the report script and its unedited output" and contains neither. | Self-evident on both pages. | **Accepted in part.** The excerpt problem is fixed (§11.1 B) and the missing output and 83-paper list are fixed — both are now published below, in full and unedited. **Rejected:** moving the fold section off the content page. Other pages on this site carry a "Reproducing these figures" section, a reader who wants to check a family assignment should not have to change page, and the section already sits after the family table it explains. **Also rejected:** linking a repository — there is no public one, and the honest fix was to publish the artefacts here rather than to point at something nobody can open. |
+| 10 | should-fix | §9 was missing (headings ran 1–8, 10, 11), and the page overstates its own completeness — "Rejected: not recorded" under a heading whose purpose is recording rejections. | Self-evident. | **Accepted.** It was a numbering slip, not a dropped section: renumbered to 1–10, with this pass as §11. The rejected-sources gap is now named as a gap, and this pass records its own rejections in this table. |
+| 11 | nit | "More than four in five of the papers measuring fingerprinting do not say whether they were detectable" mis-scopes 17.5%, which is over the 57 crawling papers and is about stating headless, not about detectability. | Report line: headless is 10 of 57. | **Accepted**, rewritten with the right denominator and the distinction between reporting the mode and reporting detectability. |
+| 12 | nit | "Usability of fingerprinting defences is invisible here" — the 83 include a PETS 2026 perception study (this page's own §5 folds its tuples into browser fingerprinting) and Berke et al. | §5 of this page. | **Accepted**: "invisible" → "thin", with both papers named. |
+| 13 | nit | The surface table is undated, so a reader cannot tell which rows are historical surfaces. | ''navigator.plugins'': WHATWG HTML Standard now specifies a fixed ''PluginArray'' (fetched 2026-09-03; the spec text names "Chromium PDF Viewer", "WebKit built-in PDF"). Battery API: Bugzilla **1313580** "Remove web content access to Battery API", queried via the REST API on 2026-09-03 — ''status: VERIFIED'', ''resolution: FIXED'', ''cf_status_firefox52: verified''. | **Accepted.** A ''%%<WRAP important>%%'' after the surface table dates the rows. This is the single fix most aligned with the currency requirement, and it is the one the corpus alone could never have produced. |
+| 14 | nit | "the most consequential change since GDPR … a web in which the largest ad platform contractually forbade the technique" — the policy bound Google's advertising customers, not the web. | The ICO source already on the page describes a change to Google's own advertising-product policy. | **Accepted**, hedged and scoped. |
+| 15 | nit | PhishPrint is named without a citation though ''acharya2021_phishprint'' is in the bibliography. | ''literature:bibliography'' line 5703. | **Accepted**, cited. |
+==== 11.3 What this pass could not establish ====
+  * **Whether the "59 tuples in 31 papers" counterfactual can be reproduced.** The content page says the pre-2026-08-12 fold would have left that residue on the current corpus. That figure is recorded in §5 but **no published artefact can regenerate it**, because the old fold was overwritten rather than kept. It is the one number on the page a reader cannot check. Keeping a dated copy of a fold before widening it would close this; it was not done.
+  * **Whether the 103 technique-only exclusions are correctly excluded.** Unchanged from §8 — still never audited paper by paper.
+  * **Whether the surface-dating box is complete.** Four surfaces were dated against primary sources. WebRTC, ''AudioContext'' and the extension-enumeration channels were not re-checked against current browser behaviour, and a reader should not read their absence from the box as "unchanged".
+  * **The 2025 policy-reversal section was not re-verified** in this pass. §7's warning still stands: it is the part of the page most likely to be stale first.
+==== 11.4 External sources: used, and rejected ====
+^ Source ^ Used for ^ Verification ^
+| ''mozilla-services/shavar-prod-lists'', ''disconnect-blacklist.json'' | Disconnect's two fingerprinting categories and their sizes | Fetched raw from GitHub 2026-09-03, parsed, category counts printed |
+| ''duckduckgo/tracker-radar'', ''docs/DATA_MODEL.md'' + ''README.md'' | the ''fingerprinting'' score, the ''apis'' field, the licence | Fetched raw 2026-09-03; four domain files sampled to confirm the score is populated |
+| WHATWG HTML Standard, § PDF viewing support | ''navigator.plugins'' is a fixed list | Fetched the spec page 2026-09-03 and read the mandated ''PluginArray'' names out of it |
+| Bugzilla 1313580 | Battery API removed from Firefox web content | Queried the Bugzilla REST API 2026-09-03; status and ''cf_status_firefox52'' read from the JSON, not from a summary |
+| ''wspr-ncsu/visiblev8'' README + GitHub API | what VisibleV8 is; that it is maintained | README fetched raw and quoted verbatim; repository metadata (''pushed_at 2026-07-15'', ''archived: false'') and release tags read from the API |
+| ''programming:crawler:openwpm'' on this wiki | the Firefox pin and the ''collection_fingerprinting'' defaults | A sibling page that cites the repository at a named commit; not re-derived from the repository here |
+**Rejected:**
+  * **The Web Almanac / HTTP Archive fingerprinting figure** the reviewer suggested adding ("FingerprintJS on 0.57% of sites, 2024 chapter"). Not verified against the chapter, so **not published**. HTTP Archive is a legitimate dataset for this page and a future pass should add it — with the figure read out of the chapter, not out of a review summary.
+  * **The reviewer's "score 0–3" for Tracker Radar's ''fingerprinting'' field.** The documentation says only "likelihood"; values 1, 2 and 3 were observed. The page states what was observed and does not assert the range.
+  * **The reviewer's claim that VisibleV8 is "the instrumentation behind the 2025 CCS function-level V8-bytecode detector".** No such paper was located in the five VisibleV8 users in this population; the page names the 2021, 2023 and 2026 papers that were verified instead.
+==== 11.5 The report script, and its real unedited output ====
+Re-run on 2026-09-03 against ''data/extract/run1'' (5,859 papers) after the residue fix. This is the complete output, not an extract — including the tool residue (§11.2 #4) and the 83-paper list (§11.2 #9), both of which the page previously promised and did not publish.
+<code>
+$ node scripts/report_fingerprinting.mjs
+</code>
+<file text fp_report_2026-09-03.txt>
+corpus                                        5859 papers
+detection fired at all                        5655 papers
+phenomenon names a fingerprint                280 papers
+only the technique does (excluded)             103 papers
+## What "fingerprinting" means in this corpus
+distinct free-text strings across 280 papers.
+What was being fingerprinted               Papers  Share of 280  Distinct strings
+-----------------------------------------  ------  ------------  ----------------
+Website / traffic fingerprinting           105     37.5%         136
+Browser fingerprinting (the web platform)  83      29.6%         156
+Remote host & network-stack ID             38      13.6%         53
+Physical-object & hardware forensics       28      10.0%         54
+Human biometrics                           16      5.7%          24
+Off-web device identification              13      4.6%          18
+Media content identification               8       2.9%          11
+unmapped                                   6       2.1%          7
+Analysis-environment evasion               2       0.7%          4
+LLM / model identification                 1       0.4%          4
+Shares exceed 100%: 18 papers (6.4%) fall in more than one family.
+Tuples: 544 total — 441 resolved by the string, 76 by paper context (property-only strings), 7 unresolved.
+of 544 strings name only a property of a fingerprint, not its subject.
+## Unmapped residue (printed in full)
+tuples, 6 papers
+  CCS/2011/bitshred-feature-hashing-malware-for-scalable-triage-and-semantic-analysis
+      Fingerprint similarity approximation
+  USENIX/2016/an-empirical-study-of-textual-key-fingerprint-representations
+      Fingerprint comparison speed
+  USENIX/2016/identifying-and-characterizing-sybils-in-the-tor-network
+      Fingerprint changes
+  CCS/2019/privacy-aspects-and-subliminal-channels-in-zcash
+      value fingerprinting
+  CCS/2019/privacy-aspects-and-subliminal-channels-in-zcash
+      fingerprint survival
+  USENIX/2025/shadows-in-cipher-spaces-exploiting-tweak-repetition-in-hardware-memory-encrypti
+      Application page fingerprints
+  NDSS/2026/snpeek-side-channel-analysis-for-privacy-applications-on-confidential-vms
+      PHH fingerprinting leakage
+## The browser-fingerprinting subset: 83 papers
+Property                                 Papers  Share of 83
+---------------------------------------  ------  -----------
+measured the web platform                78      94.0%
+measured mobile                          16      19.3%
+ran an automated web crawl               57      68.7%
+empirical (measured the world as it is)  82      98.8%
+proposes a system, attack or defence     54      65.1%
+reanalyses an existing dataset           36      43.4%
+recruited human participants             19      22.9%
+assessed a law                           10      12.0%
+released an artifact link                77      92.8%
+## Over time, and by venue
+Bucket      Corpus papers  Browser FP  Per 1,000 papers  Website/traffic FP  Per 1,000
+----------  -------------  ----------  ----------------  ------------------  ---------
+–2013   511            3           5.9               7                   13.7
+–2017   769            10          13.0              11                  14.3
+–2021   1439           23          16.0              25                  17.4
+–2024   1955           29          14.8              35                  17.9
+–2026*  1185           18          15.2              27                  22.8
+Venue    Corpus papers  Browser FP  Share of venue
+-------  -------------  ----------  --------------
+WWW      843            17          2.0%
+PETS     510            17          3.3%
+USENIX   1410           14          1.0%
+CCS      990            11          1.1%
+NDSS     701            9           1.3%
+IEEE-SP  767            9           1.2%
+IMC      638            6           0.9%
+## Which surfaces and questions the browser-FP papers name
+Surface or question                             Papers  Share of 83
+----------------------------------------------  ------  -----------
+Uniqueness / entropy / anonymity set            34      41.0%
+Fingerprinting scripts as the detection target  33      39.8%
+Canvas                                          23      27.7%
+Stability & evolution over time                 14      16.9%
+Fonts                                           11      13.3%
+Audio (AudioContext / Web Audio)                10      12.0%
+Browser extensions as a signal                  10      12.0%
+Defences & their resistance                     9       10.8%
+WebGL / GPU                                     8       9.6%
+navigator / User-Agent / plugins                6       7.2%
+WebRTC                                          6       7.2%
+Timing & clock                                  4       4.8%
+Screen & display                                3       3.6%
+Device sensors (motion, battery)                2       2.4%
+TLS / network-layer signal                      0       0.0%
+A paper can name several; shares do not sum. Rankings, not precise figures — the underlying field is free text.
+## Tools and resources, folded across spellings
+Of 83 browser-FP papers, tools used or produced (`usedOrMentioned` filtered):
+Tool family                       Papers  Share of 83
+--------------------------------  ------  -----------
+EasyList / EasyPrivacy            19      22.9%
+OpenWPM                           15      18.1%
+FingerprintJS                     14      16.9%
+Disconnect list                   13      15.7%
+Brave                             7       8.4%
+Tracker Radar Collector           6       7.2%
+uBlock Origin                     6       7.2%
+FP-Inspector                      4       4.8%
+AmIUnique                         4       4.8%
+Panopticlick / Cover Your Tracks  2       2.4%
+Tor Browser                       2       2.4%
+puppeteer-extra-plugin-stealth    2       2.4%
+FP-Scanner / FP-Collect           1       1.2%
+FPMON                             1       1.2%
+Unmapped tool residue: 398 distinct names not in TOOL_FAMILIES, of which 55 are named by 2 or more of the 83 papers. Those 55, in full:
+Unmapped tool name        Papers
+------------------------  ------
+chrome                    22
+firefox                   13
+selenium                  13
+puppeteer                 12
+esprima                   6
+python                    6
+chrome devtools protocol  5
+chromium                  5
+docker                    5
+random forest             5
+virustotal                5
+visiblev8                 5
+aws                       4
+cookiepedia               4
+edge                      4
+mitmproxy                 4
+scikit-learn              4
+browserstack              3
+frida                     3
+opera                     3
+playwright                3
+safari                    3
+adblockparser             2
+adguard                   2
+appium                    2
+aws ec2                   2
+consent-o-matic           2
+ghostery                  2
+grid'5000                 2
+jsbeautifier              2
+k-nearest neighbors       2
+kraken                    2
+lighthouse                2
+logistic regression       2
+louvain method            2
+mcafee                    2
+mystique                  2
+postgresql                2
+prolific                  2
+python requests           2
+qualtrics                 2
+random forest (custom)    2
+redis                     2
+samsung remote test lab   2
+scikit-optimize           2
+selenium webdriver        2
+semi-hard triplet loss    2
+sunspider                 2
+tranco                    2
+v8                        2
+wayback machine           2
+webgl                     2
+webshrinker               2
+xhound                    2
+xvfb                      2
+The remaining 343 appear in one paper each and are listed below.
+a-miu nique, adblock plus, adblock-rs, adblockplus, adguard chinese ads filter, adguard's base filter, adguard's mobile ads filter, adguard's tracking protection filter, adobe flash, adservers list, amazon ec2, amazon mechanical turk, anaeng, android app, android debug bridge, android debug bridge (adb), android emulator, android studio, angular 11.0.4, anubis, apache, apidiff, apktool, appbrain, aristaeus, audio-fingerprint test page, autoconsent, aware, aws lightsail, bad-asn-list, bag-of-words model (custom), better tracker blocker, bfad, binomial cdf, blindelephant, bluecava, botd, browsec vpn, browser polygraph, browserleaks, browserling, browserprint, burp suite, caniuse, cantina+, canvas defender, carnus, carnus honeypage, celery, chameleon, charles web-proxy, chatgpt, chi-squared tests of independence, chrome developer tools, chrome devtools protocol (cdp), chrome extension, chrome remote debugging protocol, chrome-remoteinterface, chromevox, chromium 117.0, chromium browser extension, chromium profiler, cisco talos, clientjs, clientsim, cloud firebase, cloudflare domain intelligence api, coinbase, convolutional neural network, convolutional neural network (cnn), cookie database, cookiebot, cookiegraph, crossbrowsertesting, crt.sh, cryptofp, css-inline, custom android app, custom chrome browser extension, custom chrome extension, custom chromium-based crawler, custom crawler, custom differential grammar-based fuzzer, custom eme javascript script, custom fingerprinting analysis pipeline, custom fingerprinting heuristics, custom imss identifier, custom javascript integration, custom parsers, custom regular-expression library, custom script, custom template-attack framework, d rawn a part, data collection browser extensions, datadome, dbscan, decision tree, deep convolutional neural network, didomi, doccano, docker compose, dolphin, drawnapart, droidbot, duckduckgo autoconsent, duckduckgo entity map, duckduckgo web tracker blocklist, elasticsearch, embedding model with glove, eme track, ermes proxy, estraverse, exthouse, extremely randomized trees, f5 bot and fraud detection/defense platform, fakeium, fanboy, fasttext, fathom, feed-forward neural network (custom), fingeprintls (fptls), fingerprint analysis tool, fingerprinting script (custom), firefox for android, flask, floc, foxhound, fp-extractor, fp-fed, fp-rainbow, fp-spoofer, fpcentral, fpserver, fptrace, fritap, fully connected neural network (multilayer perceptron classifier), google chrome, google chrome m114, google closure compiler, google cloud vision api, google compute engine, google natural language processing api, google public dns, google safe browsing, google search, google translate, google-play-scraper, google's natural language api, gridsearchcv, guimark, gummy browser, hostinger, hosttracker, html5 web cryptography api, https everywhere, hugging face's trainer, huggingface trainer api, ibm watson speech to text, ida pro, ie, imagehash, infomeasure, instrumented chromium, instrumented firefox, ip2location lite ip-asn database, ipinfo, ipinfo dataset, ipregistry, ipstack, isolation forest, jaccard similarity, jadx, javascript, javascript obfuscator tool, javascript template attacks, javascript-obfuscator, jaws, jax/tensorflow, jeb, jenks natural breaks, js-beautify, jsbench, jshelter, jsnice, jsrestrictor, jstap, jupyter notebooks, justdomains, keras, kmeans clustering algorithm, kubernetes, lda, leakage detector, let's encrypt, levenshtein distance, lightbeam, local dns stub resolver, local stand-in server, longest common subsequence (lcs), magisk, maxmind geoip, maxmind geolite2, maxmind minfraud api, mcafee smartfilter internet database, mdn web api, mdn web apis, mdn web docs, memento api, metamask, metamask automator, microsoft smartscreen, minfraud, modernizr, modheader, mongodb, moss, mozilla developer network (mdn), mozilla firefox, namecheap, network advertising initiative (nai), nginx, nltk, nodejs, nordvpn, noscript, nrc lexicon, nvda, ocr library, omnicrawl, onetrust, ontheflydom, pagegraph crawler, pandas, panoptichrome, paraphrase-multilingual-mpnet-base-v2, performance api, performancenavigationtiming api, peter-lowe's blocklist, petinspector, petportal, phash, phishprint, php, polisis, prebid.js, prefixlsh, principal component analysis, priv-accept, privacy pioneer, privaricator, prototype analyzer, puppeteercluster, pyautogui, pydnsbl, pysimilar, python difflib.sequencematcher, python's cookie jar, pytorch, pyvirtualdisplay, quantcast, raccoon app downloader, raider, random forest (rf), random forest ensemble classifier, ratcliff-obershelp, ratcliff-obershelp algorithm, ray tune, recaptcha v2, regular expressions, relevantid, request interceptor, requests, resnet50, resource timing api, routeviews, safari content blocker api, safedns website categorization service, samsung internet browser, savitzky-golay filter, scikit-learn resample, scrapy, seclists, selenium chromedriver, sentencetransformers, shap, simhash, spacy, spartacus, sqlite, standard scaler, strace, stress, stylisticfp, support vector machines (svm), svm, svm (custom), tailscale, talkback, tcpdump, tensorflow, tensorflow.js, texttiling, tf-idf, three.js, time-series k-means with dynamic time warping, tinybert, tinyproxy, tldextract, tor, tor proxy list, trackerlist, trackingexcavator, transformer, treeinterpreter, trendmicro, ua parser, uiharvester, uniquemachine, user agent switcher, user timing api, virtualbox, virustotal api, visiblejs, voiceover, wallet simulator, web idl, webrate, webviewtracer, webxray, weka, wepawet, wespoofed, wget, whatweb, whoer, whois, whotracks.me, whotracksme, widevine integration platform, windows sysinternals, word2vec, xgboost, yandex browser
+Classification resources named by the same 83 papers:
+Resource family                   Papers
+--------------------------------  ------
+EasyList / EasyPrivacy            17
+Disconnect list                   13
+FP-Inspector                      5
+FingerprintJS                     4
+Tracker Radar Collector           2
+OpenWPM                           1
+Panopticlick / Cover Your Tracks  1
+FP-Scanner / FP-Collect           1
+Brave                             1
+uBlock Origin                     1
+FPMON                             1
+Folding check — FingerprintJS appears under 7 spellings here: fingerprintjs, fingerprintjs2, FingerprintJS2, Fingerprint2, FingerprintJS, FPJS Pro, Fingerprintjs2 (FPJS2)
+## Crawl configuration: 57 of 83 ran a crawl, 57 recorded a configuration
+crawlConfig field  States a value  Share of 57  All 1080 crawling papers
+-----------------  --------------  -----------  ------------------------
+statefulness       21              36.8%        21.9%
+interactionDepth   52              91.2%        77.9%
+consentAction      39              68.4%        54.2%
+headless           10              17.5%        13.0%
+authentication     55              96.5%        72.1%
+browsers           44              77.2%        49.0%
+Browser driven (folded)                Papers
+-------------------------------------  ------
+Chrome / Chromium                      29
+Firefox (incl. versions)               17
+IE                                     1
+Opera                                  1
+default browser                        1
+Spartacus browser                      1
+Brave                                  1
+Tor Browser                            1
+DuckDuckGo                             1
+Ghostery                               1
+Android WebView                        1
+basic HTTP fetcher                     1
+Puppeteer/TRC desktop browser          1
+Puppeteer/TRC emulated mobile browser  1
+## Prevalence claims stated by browser-FP papers
+of 83 state a headline prevalence for a browser-FP phenomenon (97.6%).
+NDSS/host-fingerprinting-and-tracking-on-the-web-privacy-and-security-implications
+      Browser fingerprint entropy — UA+IP entropy was 20.29 bits
+IEEE-SP/cookieless-monster-exploring-the-ecosystem-of-web-based-device-fingerprinting
+      commercial web fingerprinting adoption — 40 sites (0.4% of the Alexa top 10,000)
+IEEE-SP/cookieless-monster-exploring-the-ecosystem-of-web-based-device-fingerprinting
+      browser-family fingerprinting — Each browser family had at least two unique features
+IEEE-SP/cookieless-monster-exploring-the-ecosystem-of-web-based-device-fingerprinting
+      browser-version fingerprinting — 71 sets identified specific Google Chrome versions
+IEEE-SP/cookieless-monster-exploring-the-ecosystem-of-web-based-device-fingerprinting
+      system fingerprinting plugins — Two of three providers searched for and loaded special plugins
+CCS/the-web-never-forgets-persistent-tracking-mechanisms-in-the-wild
+      canvas fingerprinting — more than 5.5% of crawled sites; 5542 of the top 100,000 sites
+WWW/privaricator-deceiving-fingerprinters-with-little-white-lies
+      Fingerprintable browser-property usage — 82.3% of scripts had 0 offsetHeight accesses; 1.87% had more than 50
+WWW/privaricator-deceiving-fingerprinters-with-little-white-lies
+      Fingerprinting resistance — 96.32% BlueCava; 78.36% fingerprintjs; 37.83% PetPortal; 78.81% Coinbase
+CCS/online-tracking-a-1-million-site-measurement-and-analysis
+      canvas fingerprinting — 14,371 sites (1.6%); 98.2% came from third-party scripts.
+CCS/online-tracking-a-1-million-site-measurement-and-analysis
+      canvas font fingerprinting — 3,250 first-party sites; 2.5% of the top 1,000.
+CCS/online-tracking-a-1-million-site-measurement-and-analysis
+      AudioContext fingerprinting — Two fingerprinting scripts utilizing the API were discovered.
+CCS/online-tracking-a-1-million-site-measurement-and-analysis
+      Battery API fingerprinting — Two fingerprinting scripts were identified.
+PETS/near-optimal-fingerprinting-with-constraints
+      Targeted fingerprint uniqueness — 96.15% of users had anonymity sets of at most 3 at s = 50
+PETS/near-optimal-fingerprinting-with-constraints
+      Font fingerprint uniqueness — 13,155 users (30.13%) were uniquely identified using 10 fonts
+PETS/near-optimal-fingerprinting-with-constraints
+      General fingerprint uniqueness — 19.02% of users were unique with 10 fonts and available screen size
+USENIX/internet-jones-and-the-raiders-of-the-lost-trackers-an-archaeological-study-of-w
+      fingerprint-related API use — use of two or more fingerprinting APIs rose over time
+IEEE-SP/beauty-and-the-beast-diverting-modern-web-browsers-to-build-unique-browser-finge
+      browser fingerprint uniqueness — 89.4% of 118,934 fingerprints
+IEEE-SP/beauty-and-the-beast-diverting-modern-web-browsers-to-build-unique-browser-finge
+      mobile browser fingerprinting — 81% of 13,105 mobile fingerprints
+IEEE-SP/beauty-and-the-beast-diverting-modern-web-browsers-to-build-unique-browser-finge
+      desktop browser fingerprinting — 90% of desktop fingerprints
+IEEE-SP/beauty-and-the-beast-diverting-modern-web-browsers-to-build-unique-browser-finge
+      canvas fingerprinting — 0.491 overall
+IEEE-SP/cloak-of-visibility-detecting-when-machines-browse-a-different-web
+      cloaking fingerprinting techniques — JavaScript support: 22.4% of search and 49.6% of ads
+USENIX/a-privacy-analysis-of-cross-device-tracking
+      device fingerprint uniqueness — 3 duplicate mobile fingerprints and 1 duplicate desktop fingerprint
+USENIX/extension-breakdown-security-analysis-of-browsers-extension-resources-control-po
+      extension-set device fingerprinting — 116 of 204 users, or 56.86%, had unique extension sets.
+IEEE-SP/xhound-quantifying-the-fingerprintability-of-browser-extensions
+      browser-extension fingerprintability — At least 9.2% on arbitrary domains and more than 16.6% on popular URLs among the top 10,000 Chrome extensions
+IEEE-SP/xhound-quantifying-the-fingerprintability-of-browser-extensions
+      user extension fingerprintability — 14.1% uniquely identifiable; 19.4% share fingerprints with 2–20 users
+IEEE-SP/xhound-quantifying-the-fingerprintability-of-browser-extensions
+      persistent extension fingerprintability — 88.0% of originally detectable extensions remained detectable
+IEEE-SP/xhound-quantifying-the-fingerprintability-of-browser-extensions
+      Firefox extension fingerprintability — 16% on at least one URL and 7.3% on any domain
+IEEE-SP/xhound-quantifying-the-fingerprintability-of-browser-extensions
+      extension-fingerprinting script performance — Less than one second with up to 20 installed extensions
+CCS/clock-around-the-clock-time-based-device-fingerprinting
+      Web fingerprint discrimination — CryptoFP achieved around 45% exactly unique fingerprints among 300 users
+CCS/the-webs-sixth-sense-a-study-of-scripts-accessing-smartphone-sensors
+      browser fingerprinting — 62.7% of motion-sensor scripts also fingerprinted
+IEEE-SP/fp-stalker-tracking-browser-fingerprint-evolutions
+      browser fingerprint evolution — 50% of browser instances changed fingerprints in less than 5 days; 80% in less than 10 days
+IEEE-SP/fp-stalker-tracking-browser-fingerprint-evolutions
+      browser fingerprint linkability — 54.48 consecutive days on average for the hybrid variant
+IMC/how-tracking-companies-circumvented-ad-blockers-using-websockets
+      browser fingerprinting — approximately 3.4% of WebSockets
+USENIX/fp-scanner-the-privacy-implications-of-browser-fingerprint-inconsistencies
+      browser fingerprint inconsistencies — FP-Scanner reaches accuracy 1.0 overall
+WWW/hiding-in-the-crowd-an-analysis-of-the-effectiveness-of-browser-fingerprinting-a
+      unique browser fingerprints — 33.6% of fingerprints were unique; 35.7% on personal computers and 18.5% on mobile devices
+WWW/hiding-in-the-crowd-an-analysis-of-the-effectiveness-of-browser-fingerprinting-a
+      fingerprint fragility — More than 85% of desktop/laptop fingerprints became unique after one change
+IMC/tales-from-the-porn-a-comprehensive-privacy-analysis-of-the-web-porn-ecosystem
+      canvas fingerprinting — 245 scripts in 315 websites; third-party services appeared in 4% of sites
+WWW/evaluating-anti-fingerprinting-privacy-enhancing-technologies
+      Fingerprint trackability — Tor Browser Bundle revealed under 3 bits of entropy in the hybrid evaluation
+WWW/unnecessarily-identifiable-quantifying-the-fingerprintability-of-browser-extensi
+      bloat-related fingerprintable side effects — 3,320 extensions, or 5.7% of all extensions
+WWW/unnecessarily-identifiable-quantifying-the-fingerprintability-of-browser-extensi
+      bloat-only fingerprintability — approximately 62% of fingerprintable extensions had at least one bloat-related side effect
+WWW/unnecessarily-identifiable-quantifying-the-fingerprintability-of-browser-extensi
+      unique bloat fingerprints — 61.2% had totally distinct fingerprints
+NDSS/carnus-exploring-the-privacy-threats-of-browser-extension-fingerprinting
+      web-accessible-resource extension fingerprinting — 25,866 extensions detected
+NDSS/carnus-exploring-the-privacy-threats-of-browser-extension-fingerprinting
+      DOM-based extension fingerprinting — 5,793 extensions detected
+NDSS/carnus-exploring-the-privacy-threats-of-browser-extension-fingerprinting
+      intra-browser communication fingerprinting — 450 extensions detected
+NDSS/carnus-exploring-the-privacy-threats-of-browser-extension-fingerprinting
+      extension HTTP-request fingerprinting — 859 extensions detected
+NDSS/carnus-exploring-the-privacy-threats-of-browser-extension-fingerprinting
+      fingerprintability under CloakX — 83.6%–87.92% remain identifiable
+PETS/long-term-observation-on-browser-fingerprinting-users-trackability-and-perspecti
+      browser fingerprint trackability — 64.6%-94.5% of users were trackable across evaluated feature sets
+PETS/long-term-observation-on-browser-fingerprinting-users-trackability-and-perspecti
+      fingerprint stability — 10.7-11.9 weeks when optimizing for stability
+PETS/long-term-observation-on-browser-fingerprinting-users-trackability-and-perspecti
+      user perceptions of fingerprinting — 99.1% agreed websites can recognize users
+WWW/beyond-the-front-page-measuring-third-party-dynamics-in-the-field
+      device fingerprinting — Device fingerprinting increased by 25% on subsites; 0.15% of landing pages contained the script.
+WWW/the-representativeness-of-automated-web-crawls-as-a-surrogate-for-human-browsing
+      browser fingerprinting — Around 12% of Trexa sites, mostly Canvas fingerprinting.
+IEEE-SP/fingerprinting-the-fingerprinters-learning-to-detect-browser-fingerprinting-beha
+      browser fingerprinting prevalence — 10.18% of Alexa top-100K websites
+IEEE-SP/fingerprinting-the-fingerprinters-learning-to-detect-browser-fingerprinting-beha
+      fingerprinting domains — 2,349 domains on the top 100K sites
+IEEE-SP/fingerprinting-the-fingerprinters-learning-to-detect-browser-fingerprinting-beha
+      fingerprinting script classification — 99.9% accuracy, 93.8% recall, and 93.1% precision
+IEEE-SP/fingerprinting-the-fingerprinters-learning-to-detect-browser-fingerprinting-beha
+      JavaScript API use by fingerprinters — 25 API-keyword clusters extracted
+PETS/ml-cb-machine-learning-canvas-block
+      canvas fingerprinting — 21,395 of 280,148 files leveraged canvas fingerprinting
+PETS/ml-cb-machine-learning-canvas-block
+      canvas fingerprinting — 11% at the time of the 2018 scrape
+NDSS/tales-of-favicons-and-caches-persistent-tracking-in-modern-browsers
+      fingerprint entropy — 16–24 bits commonly available across platforms
+PETS/unveiling-web-fingerprinting-in-the-wild-via-code-mining-and-machine-learning
+      web fingerprinting — 1.1% of JSWild scripts
+PETS/unveiling-web-fingerprinting-in-the-wild-via-code-mining-and-machine-learning
+      web fingerprinting — 4,544 scripts from 842 services
+PETS/unveiling-web-fingerprinting-in-the-wild-via-code-mining-and-machine-learning
+      unknown fingerprinters — 695 services were unknown to TrackerList
+PETS/unveiling-web-fingerprinting-in-the-wild-via-code-mining-and-machine-learning
+      fingerprinting API usage — Plug-in and MIME-type enumeration used by 85% of fingerprinting scripts
+PETS/unveiling-web-fingerprinting-in-the-wild-via-code-mining-and-machine-learning
+      canvas fingerprinting — 21% of fingerprinting scripts
+PETS/unveiling-web-fingerprinting-in-the-wild-via-code-mining-and-machine-learning
+      audio fingerprinting — 4% of fingerprinting scripts
+USENIX/fingerprinting-in-style-detecting-browser-extensions-via-injected-style-sheets
+      Unique CSS extension fingerprints — 4,446 extensions, or 3.8%, could be uniquely identified on any webpage
+USENIX/fingerprinting-in-style-detecting-browser-extensions-via-injected-style-sheets
+      CSS fingerprint persistence — 82.5% still remain fingerprintable despite their updates
+USENIX/phishprint-evading-phishing-detection-crawlers-by-prior-profiling
+      Advanced browser fingerprints — 204 distinct <F,C,W> fingerprints across 348,516 sessions
+USENIX/phishprint-evading-phishing-detection-crawlers-by-prior-profiling
+      Fingerprint specificity — Smallest comparison dataset contained 467,696 visitors
+USENIX/security-analysis-of-the-democracy-live-online-voting-system
+      Browser fingerprint collection — Online ballot return included a browser fingerprint.
+WWW/cookie-swap-party-abusing-first-party-cookies-for-web-tracking
+      fingerprinting-generated cookies — 3 cookies highly indicative of fingerprinting
+WWW/user-tracking-in-the-post-cookie-era-how-websites-bypass-gdpr-consent-to-track-u
+      browser fingerprinting — 336 websites (1.24%) in at least one consent action; 73.5% in all three cases
+IEEE-SP/good-bot-bad-bot-characterizing-automated-browsing-activity
+      application fingerprinting — 223,913 requests from 12,183 unique bot IP addresses
+CCS/im-spartacus-no-im-spartacus-proactively-protecting-users-from-phishing-by-inten
+      fingerprinting-based cloaking — 96.52% (2,831) of 2,933 phishing kits
+IMC/your-speaker-or-my-snooper-measuring-the-effectiveness-of-web-audio-browser-fing
+      Web Audio fingerprint stability — Dynamics Compressor produced one stable fingerprint for each of 2,093 users.
+IMC/your-speaker-or-my-snooper-measuring-the-effectiveness-of-web-audio-browser-fing
+      Fingerprint collation — Fingerprint match scores ranged from 0.9899 to 0.9978 for tested subset sizes.
+IMC/your-speaker-or-my-snooper-measuring-the-effectiveness-of-web-audio-browser-fing
+      Audio fingerprint diversity — The combined audio vectors yielded 95 distinct fingerprints among 2,093 users.
+IMC/your-speaker-or-my-snooper-measuring-the-effectiveness-of-web-audio-browser-fing
+      Additive fingerprinting value — Canvas + Audio increased normalized entropy by 9.6%; UA + Audio increased it by 9.7%.
+NDSS/auto-draft-242
+      browser fingerprint tracking — Tracking increased from 17.5 to 28 days for seven-day collection periods
+NDSS/auto-draft-242
+      cross-browser fingerprinting — 52.6%, 79.3%, and 89.7% for Edge, Opera, and Yandex
+NDSS/drawn-apart-a-device-identification-technique-based-on-remote-gpu-fingerprinting
+      browser fingerprint tracking — Tracking time increased from 17.5 days to 28 days for seven-day collection periods.
+PETS/fp-radar-longitudinal-measurement-and-early-detection-of-browser-fingerprinting
+      browser fingerprinting via web APIs — 1.16% in 2016 and 3.70% in 2018
+PETS/fp-radar-longitudinal-measurement-and-early-detection-of-browser-fingerprinting
+      previously unknown API fingerprinting — 44% of API keywords were still publicly undisclosed
+PETS/fp-radar-longitudinal-measurement-and-early-detection-of-browser-fingerprinting
+      fingerprinting API cluster — 61% of keywords in the fingerprinting cluster appeared in fingerprinting scripts
+PETS/my-cookie-is-a-phoenix-detection-measurement-and-lawfulness-of-cookie-respawning
+      cookie respawning with browser fingerprinting — 1,150 of 30,000 websites (3.83%)
+PETS/my-cookie-is-a-phoenix-detection-measurement-and-lawfulness-of-cookie-respawning
+      fingerprinting-feature dependency — 1,425 cookies depended on at least one studied feature
+PETS/omnicrawl-comprehensive-measurement-of-web-tracking-with-real-desktop-and-mobile
+      browser fingerprinting — Privacy-focused browsers showed 48% fewer WebGL Image and 44% fewer Canvas detections.
+PETS/omnicrawl-comprehensive-measurement-of-web-tracking-with-real-desktop-and-mobile
+      fingerprinting code reuse — 277 of 1,444 scripts, or 14%, contained all fingerprintjs2 magic strings.
+USENIX/phish-in-sheeps-clothing-exploring-the-authentication-pitfalls-of-browser-finger
+      browser fingerprinting on login pages — Basic fingerprinting was widespread; advanced techniques included canvas, fonts, WebRTC, and AudioContext.
+USENIX/the-dangers-of-human-touch-fingerprinting-browser-extensions-through-user-action
+      user-triggered extension fingerprinting — 4,971 unique extensions
+USENIX/the-dangers-of-human-touch-fingerprinting-browser-extensions-through-user-action
+      user-action-only fingerprintability — ≈64% can only be detected through user-driven interactions
+WWW/beyond-bot-detection-combating-fraudulent-online-survey-takers
+      browser-fingerprint duplication — 48 MTurk worker IDs contained duplicate fingerprints and were grouped into 14 groups.
+WWW/investigating-advertisers-domain-changing-behaviors-and-their-impacts-on-ad-bloc
+      browser fingerprinting — 142 of 395 labeled RAD domains (35.9%)
+CCS/cookiegraph-understanding-and-detecting-first-party-tracking-cookies
+      Fingerprinting scripts setting cookies — 1,908 out of 20K sites
+NDSS/him-of-many-faces-characterizing-billion-scale-adversarial-and-benign-browser-fingerprints-on-commercial-websites
+      adversarial versus benign browser fingerprints — 1.6% shared, 8.1% purely adversarial, and 90.3% purely benign
+NDSS/him-of-many-faces-characterizing-billion-scale-adversarial-and-benign-browser-fingerprints-on-commercial-websites
+      fingerprint evolution — Benign fingerprints increased from near one to more than 1.5 fingerprints per account; adversarial fingerprints stayed near one
+NDSS/him-of-many-faces-characterizing-billion-scale-adversarial-and-benign-browser-fingerprints-on-commercial-websites
+      adversarial fingerprint generation tools — Scripting tools accounted for over 80% of adversarial traffic
+NDSS/him-of-many-faces-characterizing-billion-scale-adversarial-and-benign-browser-fingerprints-on-commercial-websites
+      adversarial fingerprint strategies — Block+Randomize was the most common emulated-browser strategy, at 9.0% of adversarial requests
+PETS/your-drm-can-watch-you-too-exploring-the-privacy-implications-of-browsers-mis-im
+      Android Client ID fingerprint uniqueness — The certificate chain was unique for all Android devices in the test set.
+PETS/your-drm-can-watch-you-too-exploring-the-privacy-implications-of-browsers-mis-im
+      Client ID fingerprint stability — Android devices returned the same certificate over two years and shorter intervals.
+USENIX/is-your-wallet-snitching-on-you-an-analysis-on-the-privacy-implications-of-web3
+      wallet-enhanced browser fingerprinting — 878 scripts belonging to 1,099 websites
+USENIX/one-size-does-not-fit-all-uncovering-and-exploiting-cross-platform-discrepant-ap
+      Fingerprintable APIs — 13 APIs
+USENIX/rods-with-laser-beams-understanding-browser-fingerprinting-on-phishing-pages
+      browser fingerprinting — 18.7% (319,922) of 1,709,810 sites
+USENIX/rods-with-laser-beams-understanding-browser-fingerprinting-on-phishing-pages
+      fingerprinting over time — around 12% initially to over 26.6% in the last months
+USENIX/rods-with-laser-beams-understanding-browser-fingerprinting-on-phishing-pages
+      fingerprinting intention — 32.6% content manipulation; 31.6% harvesting; 35.8% both
+USENIX/rods-with-laser-beams-understanding-browser-fingerprinting-on-phishing-pages
+      fingerprinting API usage — 890,795 calls; 78.2% collect UserAgent; 6% use canvas
+USENIX/rods-with-laser-beams-understanding-browser-fingerprinting-on-phishing-pages
+      fingerprint harvesting — 81.3% (490) of clusters sent data in requests
+USENIX/rods-with-laser-beams-understanding-browser-fingerprinting-on-phishing-pages
+      target-page fingerprint matching — 29.1% of target login pages
+WWW/automatic-discovery-of-emerging-browser-fingerprinting-techniques
+      browser fingerprinting APIs — 231 fingerprinting APIs in the first crawl; 249 across two crawls
+WWW/automatic-discovery-of-emerging-browser-fingerprinting-techniques
+      emerging fingerprinting APIs — 18 additional fingerprinting APIs
+WWW/automatic-discovery-of-emerging-browser-fingerprinting-techniques
+      fingerprinting API detection coverage — 176 of 249 fingerprinting APIs
+IEEE-SP/fashion-faux-pas-implicit-stylistic-fingerprints-for-bypassing-browsers-anti-fin
+      stylistic browser fingerprinting — StylisticFP uniquely identified 534 of 866 filtered devices.
+IEEE-SP/fashion-faux-pas-implicit-stylistic-fingerprints-for-bypassing-browsers-anti-fin
+      fingerprint stability across visits — StylisticFP failed for 41 devices; FPJS failed for 188.
+IEEE-SP/fashion-faux-pas-implicit-stylistic-fingerprints-for-bypassing-browsers-anti-fin
+      fingerprint collisions — 95 StylisticFP collisions versus 55 for FPJS
+IEEE-SP/fashion-faux-pas-implicit-stylistic-fingerprints-for-bypassing-browsers-anti-fin
+      fingerprint feature entropy — Media-2 had the highest entropy, 0.58.
+IEEE-SP/fashion-faux-pas-implicit-stylistic-fingerprints-for-bypassing-browsers-anti-fin
+      fingerprinting performance overhead — Entire page loading was under one second in 98% of runs.
+IMC/browser-polygraph-efficient-deployment-of-coarse-grained-browser-fingerprints-fo
+      fraud browser fingerprints — 75–84% recall for GoLogin, Incogniton, and Octo Browser; 67% for Sphere
+IMC/browser-polygraph-efficient-deployment-of-coarse-grained-browser-fingerprints-fo
+      fingerprint uniqueness — 0.3% of fingerprints were unique
+IMC/browser-polygraph-efficient-deployment-of-coarse-grained-browser-fingerprints-fo
+      fingerprint response overhead — 6 ms response time and 1 KB storage
+CCS/peeking-through-the-window-fingerprinting-browser-extensions-through-page-visibl
+      Global-variable fingerprinting — 1,301 extensions uniquely identified by variables
+CCS/peeking-through-the-window-fingerprinting-browser-extensions-through-page-visibl
+      Storage-side-effect fingerprinting — 391 extensions uniquely identified through storage IndexedDB behavior
+CCS/peeking-through-the-window-fingerprinting-browser-extensions-through-page-visibl
+      postMessage fingerprinting — 737 of 1,028 deterministic-message extensions sent uniquely identifying messages
+CCS/peeking-through-the-window-fingerprinting-browser-extensions-through-page-visibl
+      Chrome extension fingerprintability — 2,747 of 3,398 behaviorally active extensions uniquely fingerprintable
+CCS/peeking-through-the-window-fingerprinting-browser-extensions-through-page-visibl
+      Firefox extension fingerprintability — 572 extensions
+CCS/peeking-through-the-window-fingerprinting-browser-extensions-through-page-visibl
+      Fingerprint persistence across visits — nine visits per extension
+PETS/website-data-transparency-in-the-browser
+      browser fingerprinting — F1 score 1.00 on seven test instances
+NDSS/fp-fed-privacy-preserving-federated-detection-of-browser-fingerprinting
+      Browser fingerprinting scripts — 752 of 181,633 scripts (0.41%)
+NDSS/fp-fed-privacy-preserving-federated-detection-of-browser-fingerprinting
+      Fingerprinting scripts missed by automated crawls — Manual crawling captured 3.07× more fingerprinting scripts than automated crawlers.
+NDSS/fp-fed-privacy-preserving-federated-detection-of-browser-fingerprinting
+      Fingerprinting detector performance — 0.86 AUPRC with 1M participants and ε = 1
+PETS/fp-tracer-fine-grained-browser-fingerprinting-detection-via-taint-tracking-and-e
+      browser fingerprinting flows — 269 784 fingerprinting flows and 15 239 unique browser attribute combinations
+PETS/fp-tracer-fine-grained-browser-fingerprinting-detection-via-taint-tracking-and-e
+      entropy-based fingerprinting severity — Very High 8.08%, High 61.49%, Medium 30.94%, Low 58.99%
+PETS/fp-tracer-fine-grained-browser-fingerprinting-detection-via-taint-tracking-and-e
+      audio fingerprinting — 54 crawled domains
+PETS/fp-tracer-fine-grained-browser-fingerprinting-detection-via-taint-tracking-and-e
+      canvas fingerprinting — 6237 crawled domains
+PETS/fp-tracer-fine-grained-browser-fingerprinting-detection-via-taint-tracking-and-e
+      storage fingerprinting — storageEstimate.quota detected in 148 domains
+PETS/fp-tracer-fine-grained-browser-fingerprinting-detection-via-taint-tracking-and-e
+      fingerprinting obfuscation — 46% of transmitted attributes in the most severe category
+PETS/fp-tracer-fine-grained-browser-fingerprinting-detection-via-taint-tracking-and-e
+      collaborative fingerprinting — 38% of fingerprinters involve two or more domains
+USENIX/purl-safe-and-effective-sanitization-of-link-decoration
+      fingerprinting information sharing — 200 of 1,800 decorations were labeled ATS
+WWW/the-double-edged-sword-identifying-authentication-pages-and-their-fingerprinting
+      browser fingerprinting — 9.2% of visited pages
+WWW/the-double-edged-sword-identifying-authentication-pages-and-their-fingerprinting
+      browser fingerprinting on authentication pages — 10.2% of login and sign-up pages
+WWW/the-double-edged-sword-identifying-authentication-pages-and-their-fingerprinting
+      Canvas fingerprinting — 93.10% of fingerprinting homepages
+WWW/the-double-edged-sword-identifying-authentication-pages-and-their-fingerprinting
+      tracking fingerprinting scripts — 50.50% on login pages and 55.65% on sign-up pages
+WWW/the-double-edged-sword-identifying-authentication-pages-and-their-fingerprinting
+      same third-party fingerprinting scripts — 505 of 515 domains (98.06%)
+WWW/the-double-edged-sword-identifying-authentication-pages-and-their-fingerprinting
+      fingerprinting blocked functionality — Disruptions observed on deezer.com and hepsiburada.com
+WWW/panoptichrome-a-modern-in-browser-taint-analysis-framework
+      fingerprinting APIs — 362 APIs probably used for fingerprinting or potentially misusable; 208 were previously unreported
+IEEE-SP/targeted-and-troublesome-tracking-and-advertising-on-childrens-websites
+      browser fingerprinting — 9% in Frankfurt and 10% in NYC desktop crawls
+CCS/byte-by-byte-unmasking-browser-fingerprinting-at-the-function-level-using-v8-byt
+      Browser fingerprinting functions — 4,670 unique fingerprinting functions after cleaning and deduplication
+CCS/byte-by-byte-unmasking-browser-fingerprinting-at-the-function-level-using-v8-byt
+      Browser fingerprinting scripts — 7,539 fingerprinting scripts after URL filtering
+CCS/byte-by-byte-unmasking-browser-fingerprinting-at-the-function-level-using-v8-byt
+      Fingerprinting-function detection overhead — 158.74 ms average, a 4% increase in median page-load time
+CCS/local-frames-exploiting-inherited-origins-to-bypass-content-blockers
+      fingerprinting in local frames — 4,280 websites
+IMC/canvassing-the-fingerprinters-characterizing-canvas-fingerprinting-use-across-th
+      canvas fingerprinting — 12.7% of 16,276 popular sites; 9.9% of 17,260 tail sites
+IMC/canvassing-the-fingerprinters-characterizing-canvas-fingerprinting-use-across-th
+      fingerprinting vendor reach — Akamai reached 485 popular and 205 tail sites; FingerprintJS reached 462 and 298
+NDSS/cascading-spy-sheets-exploiting-the-complexity-of-modern-css-for-email-and-browser-fingerprinting
+      font fingerprinting — 831 of 861 (96.5 %)
+NDSS/cascading-spy-sheets-exploiting-the-complexity-of-modern-css-for-email-and-browser-fingerprinting
+      CSS arithmetic fingerprinting — 1116 OS-browser combination pairs (94.9 %)
+NDSS/cascading-spy-sheets-exploiting-the-complexity-of-modern-css-for-email-and-browser-fingerprinting
+      browser-extension fingerprinting — all 10 tested extensions
+NDSS/cascading-spy-sheets-exploiting-the-complexity-of-modern-css-for-email-and-browser-fingerprinting
+      email-client fingerprinting — 8 of 21 clients (40 %)
+PETS/how-unique-is-whose-web-browser-the-role-of-demographics-in-browser-fingerprinti
+      browser fingerprint uniqueness — 60.2% of the 8,400 browser fingerprints were unique
+PETS/how-unique-is-whose-web-browser-the-role-of-demographics-in-browser-fingerprinti
+      demographic differences in fingerprinting risk — Lower-income and older groups showed greater fingerprinting risk
+PETS/rethinking-fingerprinting-an-assessment-of-behavior-based-methods-at-scale-and-i
+      behavioral fingerprint uniqueness — At 150,000 users, average anonymity was reduced by 94.6%.
+PETS/rethinking-fingerprinting-an-assessment-of-behavior-based-methods-at-scale-and-i
+      behavioral fingerprint consistency — F1 decreased from 0.571 to 0.339 over 540 days.
+PETS/rethinking-fingerprinting-an-assessment-of-behavior-based-methods-at-scale-and-i
+      combined behavioral and browser fingerprinting — Combined fingerprinting achieved F1 scores above 0.86 with weighted voting at scale.
+USENIX/double-edged-shield-on-the-fingerprintability-of-customized-ad-blockers
+      filter-list fingerprinting — 84% of filter lists identified
+USENIX/double-edged-shield-on-the-fingerprintability-of-customized-ad-blockers
+      ad-blocker configuration fingerprinting — 0.72 for ADGUARD
+WWW/beyond-the-crawl-unmasking-browser-fingerprinting-in-real-user-interactions
+      Browser fingerprinting scripts — 695 fingerprinting scripts in real-user sessions versus 498 in the automated crawl
+WWW/beyond-the-crawl-unmasking-browser-fingerprinting-in-real-user-interactions
+      Fingerprinting API vectors — Several Audio and WebRTC APIs had Call Ratio = ∞
+WWW/beyond-the-crawl-unmasking-browser-fingerprinting-in-real-user-interactions
+      Federated fingerprinting detection — AUPRC 0.98 at ε=1, 5, and 10 versus 0.96 for the centralized automated-crawl model
+WWW/broken-access-on-the-challenges-of-screen-reader-assisted-two-factor-and-passwor
+      browser-extension fingerprinting — ChromeVox and other listed extensions accurately detected
+PETS/tracking-without-borders-studying-the-role-of-webviews-in-bridging-mobile-and-we
+      Browser and canvas fingerprinting — Canvas fingerprinting observed in six apps across fraud detection, bot identification, and cross-platform tracking
+PETS/unmasking-the-shadows-a-cross-country-study-of-online-tracking-in-illegal-movie
+      Browser fingerprinting — over 70% across all analyzed countries; nearly 90% in Australia and Brazil
+IMC/fp-inconsistent-measurement-and-analysis-of-fingerprint-inconsistencies-in-evasi
+      fingerprint attribute evasion — 97.71% for BotD and 81.66% for DataDome
+IMC/fp-inconsistent-measurement-and-analysis-of-fingerprint-inconsistencies-in-evasi
+      spatial fingerprint inconsistencies — Combined rules raised detection to 76.88% for DataDome and 70.86% for BotD
+IMC/fp-inconsistent-measurement-and-analysis-of-fingerprint-inconsistencies-in-evasi
+      temporal fingerprint inconsistencies — Combined rules reduced evasion by 48.11% against DataDome and 44.95% against BotD
+WWW/the-first-early-evidence-of-the-use-of-browser-fingerprinting-for-online-trackin
+      fingerprinting-based ad tracking — Substantial bid-value differences after changing browser fingerprints.
+WWW/the-first-early-evidence-of-the-use-of-browser-fingerprinting-for-online-trackin
+      fingerprinting-based HTTP tracking — Fake fingerprints reduced HTTP chains to 6,345 versus more than 36,000 for true fingerprints.
+WWW/the-first-early-evidence-of-the-use-of-browser-fingerprinting-for-online-trackin
+      fingerprinting under privacy regulations — Potential fingerprint-based data sharing remained under selected GDPR and CCPA settings.
+WWW/fp-rainbow-fingerprint-based-browser-configuration-identification
+      browser configuration fingerprint identification — 84.36% for individual parameters and 78.15% for multiple switches
+NDSS/cross-boundary-mobile-tracking-exploring-java-to-javascript-information-diffusion-in-webviews
+      Fingerprinting API use — over 30% of exfiltrating apps invoked more than 50 fingerprinting APIs
+NDSS/cross-boundary-mobile-tracking-exploring-java-to-javascript-information-diffusion-in-webviews
+      Canvas fingerprinting — over 742 apps used WebGLRenderingContext.getExtension; 284 used canvas.toDataURL
+PETS/an-improved-entropy-measure-for-web-browser-fingerprinting-risk
+      Browser fingerprinting risk — The combined Fingerprint had 5,973 distinct values and 60.2% unique users.
+PETS/the-masks-we-think-we-wear-privacy-threats-of-browser-extension-wallets-in-the-w
+      wallet fingerprinting — 36 of 85 wallets (42.4%), covering 29.08 million users (82%)
+PETS/analyzing-societal-awareness-and-perception-of-digital-fingerprinting-and-finger
+      Awareness of digital fingerprinting — Mean fingerprinting awareness was 2.4 versus 2.86 for cookies.
+PETS/analyzing-societal-awareness-and-perception-of-digital-fingerprinting-and-finger
+      Acceptance of fingerprinting applications — Mean acceptance was 3.90 for cybersecurity, 3.73 for law enforcement, and 3.00 for user experience.
+PETS/analyzing-societal-awareness-and-perception-of-digital-fingerprinting-and-finger
+      Qualitative concerns about fingerprinting — Own-purpose concerns were 34.0% for cybersecurity, 17.4% for law enforcement, and 42.9% for user experience.
+## Evidence quotes to spot-check (every 7th browser-FP tuple)
+  NDSS/2012/host-fingerprinting-and-tracking-on-the-web-privacy-and-security-implications  [results]
+      Browser fingerprint entropy: "In our data, UA has an entropy of 11.59 bits, while the entropy of UA+IP is 20.29 bits."
+  WWW/2015/privaricator-deceiving-fingerprinters-with-little-white-lies  [methodology]
+      Fingerprintable browser-property usage: "Overall, 82.3% of scripts have 0 accesses to offsetHeight. However, 1.87% of scripts have more than 50 accesses when visited at runtime."
+  PETS/2016/near-optimal-fingerprinting-with-constraints  [results]
+      Font fingerprint uniqueness: "Detection of 10 fonts allows unique identification of 13, 155 users in our datasets (30.13%)"
+  IEEE-SP/2018/fp-stalker-tracking-browser-fingerprint-evolutions  [results]
+      browser fingerprint evolution: "50 % of browser instances changed their fingerprints in less than 5 days, 80 % in less than 10 days"
+  WWW/2019/evaluating-anti-fingerprinting-privacy-enhancing-technologies  [results]
+      Fingerprint trackability: "It finds Tor Browser Bundle (Tor BB) to be most effective, revealing under 3 bits of entropy."
+  NDSS/2020/carnus-exploring-the-privacy-threats-of-browser-extension-fingerprinting  [methodology]
+      extension HTTP-request fingerprinting: "we obtain all resources requested by the web page and content scripts of extensions installed in the user's browser."
+  IEEE-SP/2021/fingerprinting-the-fingerprinters-learning-to-detect-browser-fingerprinting-beha  [results]
+      browser fingerprinting prevalence: "Overall, we find that more than 10.18% of top-100K websites deploy fingerprinting."
+  PETS/2021/unveiling-web-fingerprinting-in-the-wild-via-code-mining-and-machine-learning  [results]
+      web fingerprinting: "Only 1.1% of scripts is classified as actual fingerprinting."
+  USENIX/2021/fingerprinting-in-style-detecting-browser-extensions-via-injected-style-sheets  [results]
+      CSS fingerprint persistence: "after re-running our testing pipeline, 776 triggered at least one of their previously discovered triggers. In other words, 82.5% still remain fingerprintable despite their updates."
+  CCS/2022/im-spartacus-no-im-spartacus-proactively-protecting-users-from-phishing-by-inten  [results]
+      fingerprinting-based cloaking: "In total, 96.52% (2,831) out of 2,933 phishing kits contain fingerprinting-based cloaking techniques."
+  NDSS/2022/drawn-apart-a-device-identification-technique-based-on-remote-gpu-fingerprinting  [results]
+      browser fingerprint tracking: "adding D RAWN A PART to FP-S TALKER increases the tracking time, raising the median average tracking time by 10.5 days, from 17.5 days to 28 days."
+  PETS/2022/omnicrawl-comprehensive-measurement-of-web-tracking-with-real-desktop-and-mobile  [results]
+      fingerprinting code reuse: "out of the 1444 unique scripts that match at least one fingerprinting category, 277 (14%) contain all the magic strings of fingerprintjs2"
+  NDSS/2023/him-of-many-faces-characterizing-billion-scale-adversarial-and-benign-browser-fingerprints-on-commercial-websites  [results]
+      adversarial versus benign browser fingerprints: "Only 1.6% of unique fingerprints are shared between adversarial and benign; by contrast, 8.1% are purely adversarial and 90.3% purely benign."
+  USENIX/2023/is-your-wallet-snitching-on-you-an-analysis-on-the-privacy-implications-of-web3  [results]
+      wallet-enhanced browser fingerprinting: "we find that 878 scripts (79%) belonging to 1, 099 websites (83%) engage in browser fingerprinting and leverage wallet information to enhance the fingerprints they generate."
+  USENIX/2023/rods-with-laser-beams-understanding-browser-fingerprinting-on-phishing-pages  [results]
+      fingerprinting intention: "104, 279 (32.6%) ... use it for content manipulation, and 101, 030 (31.6%) ... shares them ... Finally, 114, 613 (35.8%) use fingerprinting for both intentions."
+  CCS/2024/peeking-through-the-window-fingerprinting-browser-extensions-through-page-visibl  [results]
+      Chrome extension fingerprintability: "Out of those, we can uniquely identify 2,747."
+  NDSS/2024/fp-fed-privacy-preserving-federated-detection-of-browser-fingerprinting  [dataset]
+      Browser fingerprinting scripts: "According to our high-precision ground-truth heuristic, 752 out of the 181,633 scripts (0.41%) are fingerprinting."
+  PETS/2024/fp-tracer-fine-grained-browser-fingerprinting-detection-via-taint-tracking-and-e  [results]
+      storage fingerprinting: "We detected storageEstimate.quota in 148 domains."
+  WWW/2024/the-double-edged-sword-identifying-authentication-pages-and-their-fingerprinting  [results]
+      tracking fingerprinting scripts: "home pages have the highest rate of tracking at 61.46%, compared to login pages at 50.50% and sign-up pages at 55.65%."
+  CCS/2025/local-frames-exploiting-inherited-origins-to-bypass-content-blockers  [results]
+      fingerprinting in local frames: "Approximately a third (4,280) of all websites containing local frames use them to perform fingerprinting, a clear privacy concern."
+  PETS/2025/how-unique-is-whose-web-browser-the-role-of-demographics-in-browser-fingerprinti  [results]
+      browser fingerprint uniqueness: "Approximately 60% of users in our dataset have a unique overall Fingerprint."
+  WWW/2025/beyond-the-crawl-unmasking-browser-fingerprinting-in-real-user-interactions  [results]
+      Browser fingerprinting scripts: "out of these collected scripts, 695 were found to be fingerprinting in the real user browsing sessions, compared to only 498 in the automated crawl."
+  PETS/2026/analyzing-societal-awareness-and-perception-of-digital-fingerprinting-and-finger  [results]
+      Awareness of digital fingerprinting: "We observed an overall mean awareness level of 2.86 ... and 2.4 ... respectively (min: 1 - max: 4)."
+  IEEE-SP/2016/beauty-and-the-beast-diverting-modern-web-browsers-to-build-unique-browser-finge  [results]
+      canvas fingerprinting: "Alone, as a simple rendered picture, the normalized entropy is at 0.491, putting it in the top 5 of the most discriminating attributes."
+  WWW/2025/the-first-early-evidence-of-the-use-of-browser-fingerprinting-for-online-trackin  [results]
+      fingerprinting-based ad tracking: "When comparing "True FP True IP" to "Fake FP True IP," in Table 2, there are substantial differences in median and maximum bid values."
+  IEEE-SP/2017/xhound-quantifying-the-fingerprintability-of-browser-extensions  [results]
+      persistent extension fingerprintability: "Among the extensions that were originally fingerprintable, approximately 38% were updated and, overall, 88% of them remained fingerprintable."
+  IEEE-SP/2023/fashion-faux-pas-implicit-stylistic-fingerprints-for-bypassing-browsers-anti-fin  [evaluation]
+      fingerprinting performance overhead: "our approach is stable and the entire page's loading time is less than 1 second in 98% of the runs."
+## Legal and human-subjects angles
+of 83 browser-FP papers assess a law (corpus-wide: 402 of 5859).
+Law                                             Papers
+----------------------------------------------  ------
+GDPR                                            9
+ePrivacy Directive                              3
+CCPA                                            3
+European directives 2002/58/CE and 2009/136/CE  1
+ePrivacy directive                              1
+Digital Economy Act 2017                        1
+LGPD                                            1
+COPPA                                           1
+Digital Services Act                            1
+  WWW/2018/hiding-in-the-crowd-an-analysis-of-the-effectiveness-of-browser-fingerprinting-a  European directives 2002/58/CE and 2009/136/CE: Consent for cookies and browser fingerprinting collection
+  IMC/2019/tales-from-the-porn-a-comprehensive-privacy-analysis-of-the-web-porn-ecosystem  GDPR: Tracking, cookie consent, privacy policies, and HTTPS
+  IMC/2019/tales-from-the-porn-a-comprehensive-privacy-analysis-of-the-web-porn-ecosystem  ePrivacy directive: Cookie-consent banners before cookie use
+  IMC/2019/tales-from-the-porn-a-comprehensive-privacy-analysis-of-the-web-porn-ecosystem  Digital Economy Act 2017: Verifiable age-verification mechanisms
+  WWW/2020/beyond-the-front-page-measuring-third-party-dynamics-in-the-field  GDPR: Third-party connections to potentially inadequate countries
+  PETS/2021/unveiling-web-fingerprinting-in-the-wild-via-code-mining-and-machine-learning  GDPR: Data-collection privacy and protection procedures
+  WWW/2021/user-tracking-in-the-post-cookie-era-how-websites-bypass-gdpr-consent-to-track-u  GDPR: Tracking and personal-data processing after consent choices
+  WWW/2021/user-tracking-in-the-post-cookie-era-how-websites-bypass-gdpr-consent-to-track-u  ePrivacy Directive: Compliance of tracking mechanisms with privacy rules
+  PETS/2022/my-cookie-is-a-phoenix-detection-measurement-and-lawfulness-of-cookie-respawning  GDPR: Fairness, transparency, lawfulness, and consent for respawned cookies
+  PETS/2022/my-cookie-is-a-phoenix-detection-measurement-and-lawfulness-of-cookie-respawning  ePrivacy Directive: Consent requirements for cookie-based processing
+  NDSS/2023/him-of-many-faces-characterizing-billion-scale-adversarial-and-benign-browser-fingerprints-on-commercial-websites  GDPR: Secure processing of collected personal data
+  NDSS/2023/him-of-many-faces-characterizing-billion-scale-adversarial-and-benign-browser-fingerprints-on-commercial-websites  CCPA: Secure processing of collected personal data
+  PETS/2025/unmasking-the-shadows-a-cross-country-study-of-online-tracking-in-illegal-movie  GDPR: Differences in tracking across GDPR-country vantage points
+  PETS/2025/unmasking-the-shadows-a-cross-country-study-of-online-tracking-in-illegal-movie  CCPA: Tracking measurements from a CCPA-covered vantage point
+  PETS/2025/unmasking-the-shadows-a-cross-country-study-of-online-tracking-in-illegal-movie  LGPD: Tracking measurements from an LGPD-covered vantage point
+  WWW/2025/the-first-early-evidence-of-the-use-of-browser-fingerprinting-for-online-trackin  GDPR: Fingerprint-based data sharing after consent opt-out and opt-in.
+  WWW/2025/the-first-early-evidence-of-the-use-of-browser-fingerprinting-for-online-trackin  CCPA: Fingerprint-based data sharing after consent opt-out and opt-in.
+  IEEE-SP/2024/targeted-and-troublesome-tracking-and-advertising-on-childrens-websites  GDPR: Tracking and targeted advertising consent requirements
+  IEEE-SP/2024/targeted-and-troublesome-tracking-and-advertising-on-childrens-websites  ePrivacy Directive: Consent before tracking cookies or similar technologies
+  IEEE-SP/2024/targeted-and-troublesome-tracking-and-advertising-on-childrens-websites  COPPA: Parental consent for child-directed targeted advertising
+  IEEE-SP/2024/targeted-and-troublesome-tracking-and-advertising-on-childrens-websites  Digital Services Act: Behavioral advertising to minors
+of 83 recruited human participants:
+  USENIX/2017/a-privacy-analysis-of-cross-device-tracking  n=126  not-stated
+  USENIX/2017/extension-breakdown-security-analysis-of-browsers-extension-resources-control-po  n=204  snowball
+  CCS/2018/clock-around-the-clock-time-based-device-fingerprinting  n=565  snowball
+  IEEE-SP/2018/fp-stalker-tracking-browser-fingerprint-evolutions  n=1905  not-stated
+  PETS/2020/long-term-observation-on-browser-fingerprinting-users-trackability-and-perspecti  n=2315/1275/243  professional-network/professional-network/professional-network
+  WWW/2020/the-representativeness-of-automated-web-crawls-as-a-surrogate-for-human-browsing  n=52000  not-stated
+  PETS/2021/unveiling-web-fingerprinting-in-the-wild-via-code-mining-and-machine-learning  n=982  other-crowd-platform
+  USENIX/2021/phishprint-evading-phishing-detection-crawlers-by-prior-profiling  n=1150/1007  mechanical-turk/mechanical-turk
+  CCS/2022/im-spartacus-no-im-spartacus-proactively-protecting-users-from-phishing-by-inten  n=  not-stated
+  IMC/2022/your-speaker-or-my-snooper-measuring-the-effectiveness-of-web-audio-browser-fing  n=2064/29  mechanical-turk/not-stated
+  NDSS/2022/drawn-apart-a-device-identification-technique-based-on-remote-gpu-fingerprinting  n=  not-stated
+  WWW/2022/beyond-bot-detection-combating-fraudulent-online-survey-takers  n=289/162  social-media/mechanical-turk
+  PETS/2024/website-data-transparency-in-the-browser  n=100  prolific
+  PETS/2025/how-unique-is-whose-web-browser-the-role-of-demographics-in-browser-fingerprinti  n=12461  prolific
+  WWW/2025/beyond-the-crawl-unmasking-browser-fingerprinting-in-real-user-interactions  n=30  mechanical-turk
+  PETS/2026/analyzing-societal-awareness-and-perception-of-digital-fingerprinting-and-finger  n=261/473  other-crowd-platform/university-pool
+  IMC/2025/fp-inconsistent-measurement-and-analysis-of-fingerprint-inconsistencies-in-evasi  n=  university-pool
+  IEEE-SP/2017/xhound-quantifying-the-fingerprintability-of-browser-extensions  n=51/509/294  professional-network/mechanical-turk/university-pool
+  IEEE-SP/2023/fashion-faux-pas-implicit-stylistic-fingerprints-for-bypassing-browsers-anti-fin  n=  professional-network
+## Browser-FP papers, newest first (for citation picking)
+NDSS     cross-boundary-mobile-tracking-exploring-java-to-javascript-information-diffusion-in-webviews
+      Fingerprinting API use; Canvas fingerprinting
+PETS     an-improved-entropy-measure-for-web-browser-fingerprinting-risk
+      Browser fingerprinting risk
+PETS     the-masks-we-think-we-wear-privacy-threats-of-browser-extension-wallets-in-the-w
+      wallet fingerprinting
+PETS     analyzing-societal-awareness-and-perception-of-digital-fingerprinting-and-finger
+      Awareness of digital fingerprinting; Acceptance of fingerprinting applications; Qualitative concerns about fingerprinting
+CCS      byte-by-byte-unmasking-browser-fingerprinting-at-the-function-level-using-v8-byt
+      Browser fingerprinting functions; Browser fingerprinting scripts; Fingerprinting-function detection overhead
+CCS      local-frames-exploiting-inherited-origins-to-bypass-content-blockers
+      fingerprinting in local frames
+IMC      canvassing-the-fingerprinters-characterizing-canvas-fingerprinting-use-across-th
+      canvas fingerprinting; fingerprinting vendor reach
+IMC      fp-inconsistent-measurement-and-analysis-of-fingerprint-inconsistencies-in-evasi
+      fingerprint attribute evasion; spatial fingerprint inconsistencies; temporal fingerprint inconsistencies
+NDSS     cascading-spy-sheets-exploiting-the-complexity-of-modern-css-for-email-and-browser-fingerprinting
+      font fingerprinting; CSS arithmetic fingerprinting; browser-extension fingerprinting; email-client fingerprinting
+PETS     how-unique-is-whose-web-browser-the-role-of-demographics-in-browser-fingerprinti
+      browser fingerprint uniqueness; demographic differences in fingerprinting risk
+PETS     rethinking-fingerprinting-an-assessment-of-behavior-based-methods-at-scale-and-i
+      behavioral fingerprint uniqueness; behavioral fingerprint consistency; combined behavioral and browser fingerprinting
+PETS     tracking-without-borders-studying-the-role-of-webviews-in-bridging-mobile-and-we
+      Browser and canvas fingerprinting
+PETS     unmasking-the-shadows-a-cross-country-study-of-online-tracking-in-illegal-movie
+      Browser fingerprinting
+USENIX   double-edged-shield-on-the-fingerprintability-of-customized-ad-blockers
+      filter-list fingerprinting; ad-blocker configuration fingerprinting
+WWW      beyond-the-crawl-unmasking-browser-fingerprinting-in-real-user-interactions
+      Browser fingerprinting scripts; Fingerprinting API vectors; Federated fingerprinting detection
+WWW      broken-access-on-the-challenges-of-screen-reader-assisted-two-factor-and-passwor
+      browser-extension fingerprinting
+WWW      the-first-early-evidence-of-the-use-of-browser-fingerprinting-for-online-trackin
+      fingerprinting-based ad tracking; fingerprinting-based HTTP tracking; fingerprinting under privacy regulations
+WWW      fp-rainbow-fingerprint-based-browser-configuration-identification
+      browser configuration fingerprint identification
+CCS      peeking-through-the-window-fingerprinting-browser-extensions-through-page-visibl
+      Global-variable fingerprinting; Storage-side-effect fingerprinting; postMessage fingerprinting; Chrome extension fingerprintability; Firefox extension fingerprintability; Fingerprint persistence across visits
+IEEE-SP  targeted-and-troublesome-tracking-and-advertising-on-childrens-websites
+      browser fingerprinting
+IMC      browser-polygraph-efficient-deployment-of-coarse-grained-browser-fingerprints-fo
+      fraud browser fingerprints; fingerprint uniqueness; fingerprint response overhead
+NDSS     fp-fed-privacy-preserving-federated-detection-of-browser-fingerprinting
+      Browser fingerprinting scripts; Fingerprinting scripts missed by automated crawls; Fingerprinting detector performance
+PETS     website-data-transparency-in-the-browser
+      browser fingerprinting
+PETS     fp-tracer-fine-grained-browser-fingerprinting-detection-via-taint-tracking-and-e
+      browser fingerprinting flows; entropy-based fingerprinting severity; audio fingerprinting; canvas fingerprinting; storage fingerprinting; fingerprinting obfuscation; collaborative fingerprinting
+USENIX   purl-safe-and-effective-sanitization-of-link-decoration
+      fingerprinting information sharing
+WWW      the-double-edged-sword-identifying-authentication-pages-and-their-fingerprinting
+      browser fingerprinting; browser fingerprinting on authentication pages; Canvas fingerprinting; tracking fingerprinting scripts; same third-party fingerprinting scripts; fingerprinting blocked functionality
+WWW      panoptichrome-a-modern-in-browser-taint-analysis-framework
+      fingerprinting APIs
+CCS      cookiegraph-understanding-and-detecting-first-party-tracking-cookies
+      Fingerprinting scripts setting cookies
+IEEE-SP  fashion-faux-pas-implicit-stylistic-fingerprints-for-bypassing-browsers-anti-fin
+      stylistic browser fingerprinting; fingerprint stability across visits; fingerprint collisions; fingerprint feature entropy; fingerprinting performance overhead
+NDSS     him-of-many-faces-characterizing-billion-scale-adversarial-and-benign-browser-fingerprints-on-commercial-websites
+      adversarial versus benign browser fingerprints; fingerprint distribution differences; fingerprint evolution; adversarial fingerprint generation tools; adversarial fingerprint strategies
+PETS     your-drm-can-watch-you-too-exploring-the-privacy-implications-of-browsers-mis-im
+      Android Client ID fingerprint uniqueness; Client ID fingerprint stability
+USENIX   is-your-wallet-snitching-on-you-an-analysis-on-the-privacy-implications-of-web3
+      wallet-enhanced browser fingerprinting
+USENIX   one-size-does-not-fit-all-uncovering-and-exploiting-cross-platform-discrepant-ap
+      Fingerprintable APIs
+USENIX   rods-with-laser-beams-understanding-browser-fingerprinting-on-phishing-pages
+      browser fingerprinting; fingerprinting over time; fingerprinting intention; fingerprinting API usage; fingerprint harvesting; target-page fingerprint matching
+WWW      automatic-discovery-of-emerging-browser-fingerprinting-techniques
+      browser fingerprinting APIs; emerging fingerprinting APIs; fingerprinting API detection coverage
+CCS      privacy-limitations-of-interest-based-advertising-on-the-web-a-post-mortem-empir
+      fingerprinting-enhanced cohort unicity
+CCS      im-spartacus-no-im-spartacus-proactively-protecting-users-from-phishing-by-inten
+      fingerprinting-based cloaking
+IMC      your-speaker-or-my-snooper-measuring-the-effectiveness-of-web-audio-browser-fing
+      Web Audio fingerprint stability; Fingerprint collation; Audio fingerprint diversity; Additive fingerprinting value
+NDSS     auto-draft-242
+      browser fingerprint tracking; cross-browser fingerprinting
+NDSS     drawn-apart-a-device-identification-technique-based-on-remote-gpu-fingerprinting
+      browser fingerprint tracking
+PETS     fp-radar-longitudinal-measurement-and-early-detection-of-browser-fingerprinting
+      browser fingerprinting via web APIs; previously unknown API fingerprinting; fingerprinting API cluster
+PETS     my-cookie-is-a-phoenix-detection-measurement-and-lawfulness-of-cookie-respawning
+      cookie respawning with browser fingerprinting; fingerprinting-feature dependency
+PETS     omnicrawl-comprehensive-measurement-of-web-tracking-with-real-desktop-and-mobile
+      browser fingerprinting; fingerprinting code reuse
+USENIX   phish-in-sheeps-clothing-exploring-the-authentication-pitfalls-of-browser-finger
+      browser fingerprinting on login pages
+USENIX   the-dangers-of-human-touch-fingerprinting-browser-extensions-through-user-action
+      user-triggered extension fingerprinting; user-action-only fingerprintability
+WWW      beyond-bot-detection-combating-fraudulent-online-survey-takers
+      browser-fingerprint duplication
+WWW      investigating-advertisers-domain-changing-behaviors-and-their-impacts-on-ad-bloc
+      browser fingerprinting
+IEEE-SP  fingerprinting-the-fingerprinters-learning-to-detect-browser-fingerprinting-beha
+      browser fingerprinting prevalence; fingerprinting domains; fingerprinting script classification; JavaScript API use by fingerprinters
+IEEE-SP  good-bot-bad-bot-characterizing-automated-browsing-activity
+      application fingerprinting
+NDSS     tales-of-favicons-and-caches-persistent-tracking-in-modern-browsers
+      fingerprint entropy
+PETS     ml-cb-machine-learning-canvas-block
+      canvas fingerprinting
+PETS     unveiling-web-fingerprinting-in-the-wild-via-code-mining-and-machine-learning
+      web fingerprinting; unknown fingerprinters; fingerprinting API usage; canvas fingerprinting; audio fingerprinting
+USENIX   fingerprinting-in-style-detecting-browser-extensions-via-injected-style-sheets
+      Unique CSS extension fingerprints; CSS fingerprint persistence
+USENIX   phishprint-evading-phishing-detection-crawlers-by-prior-profiling
+      Advanced browser fingerprints; Fingerprint specificity
+USENIX   security-analysis-of-the-democracy-live-online-voting-system
+      Browser fingerprint collection
+WWW      cookie-swap-party-abusing-first-party-cookies-for-web-tracking
+      fingerprinting-generated cookies
+WWW      user-tracking-in-the-post-cookie-era-how-websites-bypass-gdpr-consent-to-track-u
+      browser fingerprinting
+NDSS     carnus-exploring-the-privacy-threats-of-browser-extension-fingerprinting
+      web-accessible-resource extension fingerprinting; DOM-based extension fingerprinting; intra-browser communication fingerprinting; extension HTTP-request fingerprinting; fingerprintability under CloakX
+PETS     long-term-observation-on-browser-fingerprinting-users-trackability-and-perspecti
+      browser fingerprint trackability; fingerprint stability; user perceptions of fingerprinting
+WWW      beyond-the-front-page-measuring-third-party-dynamics-in-the-field
+      device fingerprinting
+WWW      the-representativeness-of-automated-web-crawls-as-a-surrogate-for-human-browsing
+      browser fingerprinting
+IMC      tales-from-the-porn-a-comprehensive-privacy-analysis-of-the-web-porn-ecosystem
+      canvas fingerprinting
+WWW      evaluating-anti-fingerprinting-privacy-enhancing-technologies
+      Fingerprint trackability
+WWW      unnecessarily-identifiable-quantifying-the-fingerprintability-of-browser-extensi
+      bloat-related fingerprintable side effects; bloat-only fingerprintability; unique bloat fingerprints
+CCS      clock-around-the-clock-time-based-device-fingerprinting
+      Web fingerprint discrimination
+CCS      the-webs-sixth-sense-a-study-of-scripts-accessing-smartphone-sensors
+      browser fingerprinting
+IEEE-SP  fp-stalker-tracking-browser-fingerprint-evolutions
+      browser fingerprint evolution; browser fingerprint linkability
+IMC      how-tracking-companies-circumvented-ad-blockers-using-websockets
+      browser fingerprinting
+USENIX   fp-scanner-the-privacy-implications-of-browser-fingerprint-inconsistencies
+      browser fingerprint inconsistencies
+WWW      hiding-in-the-crowd-an-analysis-of-the-effectiveness-of-browser-fingerprinting-a
+      unique browser fingerprints; fingerprint fragility
+IEEE-SP  xhound-quantifying-the-fingerprintability-of-browser-extensions
+      browser-extension fingerprintability; user extension fingerprintability; persistent extension fingerprintability; Firefox extension fingerprintability; extension-fingerprinting script performance
+USENIX   a-privacy-analysis-of-cross-device-tracking
+      device fingerprint uniqueness
+USENIX   extension-breakdown-security-analysis-of-browsers-extension-resources-control-po
+      extension-set device fingerprinting
+CCS      online-tracking-a-1-million-site-measurement-and-analysis
+      canvas fingerprinting; canvas font fingerprinting; AudioContext fingerprinting; Battery API fingerprinting
+IEEE-SP  beauty-and-the-beast-diverting-modern-web-browsers-to-build-unique-browser-finge
+      browser fingerprint uniqueness; mobile browser fingerprinting; desktop browser fingerprinting; canvas fingerprinting; WebGL fingerprinting
+IEEE-SP  cloak-of-visibility-detecting-when-machines-browse-a-different-web
+      cloaking fingerprinting techniques
+PETS     near-optimal-fingerprinting-with-constraints
+      Targeted fingerprint uniqueness; Font fingerprint uniqueness; General fingerprint uniqueness
+USENIX   internet-jones-and-the-raiders-of-the-lost-trackers-an-archaeological-study-of-w
+      fingerprint-related API use
+WWW      privaricator-deceiving-fingerprinters-with-little-white-lies
+      Fingerprintable browser-property usage; Fingerprinting resistance
+CCS      the-web-never-forgets-persistent-tracking-mechanisms-in-the-wild
+      canvas fingerprinting
+CCS      performance-signatures-of-mobile-phone-browsers
+      mobile browser performance fingerprinting
+IEEE-SP  cookieless-monster-exploring-the-ecosystem-of-web-based-device-fingerprinting
+      commercial web fingerprinting adoption; browser-family fingerprinting; browser-version fingerprinting; system fingerprinting plugins
+NDSS     host-fingerprinting-and-tracking-on-the-web-privacy-and-security-implications
+      Browser fingerprint entropy
+</file>
+==== 11.6 Run log for this pass ====
+^ ^ ^
+| Date | 2026-09-03 |
+| Corpus | ''data/extract/run1'', 5,859 papers, unchanged since 2026-08-12 |
+| Models | Claude Opus 5 (author, verification, edits); **Claude Fable** (generic review, step 9 — the pass skipped when this page was written) |
+| Script changes | ''report_fingerprinting.mjs'': tool-fold residue now collected and printed. ''fp_fold.mjs'': header said "Two-pass design" while the code has three passes, and carried the old corpus's ''201 papers / 339 strings'' in a comment — both fixed. No figure changed. |
+| Pages saved | ''privacy:fingerprinting'' (45,805 → 69,330 bytes), this page |
+| Verified after saving | 0 red links, 35 rendered references against 35 distinct citekeys, 9 WRAP boxes, no en-dashed flags left in any ''%%<code>%%'' span |
+| Mistake caught in review | The reviewer's brief pointed it at ''out/fp_report.txt'', which is an **Aug-5 output against the 4,322-paper corpus**. It was caught before the reviewer was launched and a fresh re-run was substituted; had it not been, every figure the reviewer checked would have been checked against a stale corpus. Old output files in a working directory are a trap — this one had been sitting next to the current page for four weeks. |
+| Related item found, not fixed here | ''design:crawling_location'' was reported as an unsaved draft. It is **live** (revision of 2026-08-21) and the local draft is the //older// copy; the content page's ''%%[[Design:Crawling location]]%%'' link resolves. The stale local drafts were preserved rather than overwritten. |
+===== Amendment, 2026-09-04: citekey consolidation =====
+  * ''fouad2022my'' was one of two keys for the same paper in [[:literature:bibliography]]. The wiki-wide consolidation of 2026-09-04 (drain item ''dedup-regional-filter-lists-bibkey'') kept ''fouad2022_cookie'' and deleted the other entry.
+  * 2 citation markers on [[:privacy:fingerprinting]] were repointed to the kept key. No prose on either page changed, and no figure moved. Statements above that name the deleted key describe the state when they were written. Full query log and the invariants checked before saving: [[:provenance:literature:bibliography]].
+===== Quote-check refresh, 2026-09-21 =====
+The 2026-09-04 ''cols''-vs-PDF audit on [[:provenance:literature:corpus]] showed that 73.1% of evidence quotes that cannot be located in ''paper.cols.txt'' **are** present in an independent ''pypdf'' rendering of the same ''paper.pdf''. ''scripts/quote_check.mjs'' was patched the same day to re-check everything below threshold against that second rendering and to report a fourth verdict, **RESCUED**. §6's figure predates the patch and therefore overstated this page's quote-failure rate. Re-run, unedited first line:
+<code>
+$ node scripts/quote_check.mjs --detection 'fingerprint' --show 400
+quotes checked: 312 exact, 140 partial (>=60% of 5-word windows), 50 rescued from the PDF, 42 below threshold in both renderings, 0 with no full text on disk.
+</code>
+^ Figure ^ Was ^ Is ^ Why ^
+| quotes checked | 544 | 544 | population unchanged — the corpus has not moved |
+| exact | 312 | 312 | unchanged |
+| partial (≥60% of 5-word windows) | 140 | 140 | unchanged |
+| rescued from the PDF | — | **50** | new verdict; these were inside the old 92 |
+| below threshold | **92** | **42** (in both renderings) | 92 = 50 + 42 exactly; nothing else moved |
+**What this does and does not say.** It does not say 50 extractions were wrong and are now right — the quotes were always in the papers. It says the //stored text// could not locate them and a second rendering of the same PDF can, so counting them as quote failures was a measurement of ''decolumn.mjs'', not of the extraction. The 42 remaining are not "unsupported" either; two of them were read by hand in §6 and both are present.
+**Scope of this edit.** Only §6's quote-check figure and its surrounding prose. No fold, no report figure, no citation and no other number on either page was touched or re-derived: ''report_fingerprinting.mjs'' was **not** re-run in this pass, and every figure outside §6 stands as published on 2026-09-03. The content page's spot-check bullet was corrected in the same pass, because it described the checker as testing against ''paper.cols.txt'' alone, which is no longer what it does.
+^ Item ^ Value ^
+| Date | 2026-09-21, unsupervised |
+| Command | ''%%node scripts/quote_check.mjs --detection 'fingerprint' --show 400%%'' |
+| Artifact | ''out/qc0921/detect_fingerprint.txt'' (full run, 50 RESCUED rows and 42 below-threshold rows listed) |
+| Script changes | none — ''quote_check.mjs'' was already patched on 2026-09-04 |
+| Reviewers | one ''sonnet'' figures-vs-script pass over this page and [[:privacy:fingerprinting]] |
+| Pages saved | this page, [[:privacy:fingerprinting]] |
 [[privacy:fingerprinting|← back to the content page]] · [[literature:corpus|corpus-level provenance]]