Working log behind Public relations. Every query with its denominator, both report scripts with their unedited output, the fold rules with their residue, the quotes that were spot-checked, the external sources that were verified and the ones that were rejected, and what could not be established. Corpus-wide caveats are on Corpus and are not restated here.

Not prose. Read it if you are checking a number.

No ~~DISCUSSION~~ block, matching every existing provenance: page. Comments belong on the content page, Public relations.

Three judgement calls, each of which a reasonable person could have made differently.

• The page is about the claim, not about the interview. The brief pointed at “how to present results to journalists without overclaiming”. Two readings were available: media training (how to talk, what to wear, how to handle a hostile question), or claim discipline (which sentence travels, and what it licenses). The first is a topic on which this corpus has nothing, is well covered by any university communications office, and fails the site's “no MDN, no textbook” test. The page takes the second reading, which turned out to be measurable — see section B below. The cost of this choice is that the page gives almost no advice on the interview itself.
• A negative finding was published as a finding. Section A of the report script establishes that zero of 4,472 papers with an ethics record document their own press engagement. It would have been possible to omit this and write a purely advisory page. Publishing it instead sets the reader's expectation about how much weight the rest of the page can bear, and is the honest framing of a page with two documented cases behind it.
• A new measurement was constructed for this page. The “does a quotable sentence carry its own denominator” query (section B) is not a standard use of detection.evidence.quote and is the page's most novel and most contestable figure. It is reported with its rule, its residue in both directions, and an explicit statement of what it is a proxy for. A reasonable person would have declined to publish it, on the grounds that evidence.quote is a model's choice of sentence rather than the paper's headline sentence. It is published because the alternative was a page with no measurement on it at all, and because the limitation is stated on the content page rather than buried here.

All from node scripts/report_public_relations.mjs. Populations, in the order the page uses them:

The 4,472 denominator matches the one Notifying websites uses and differs from data/extract/OVERVIEW.md, which uses 5,118. The 646-paper gap is papers where the extractor emitted no ethics object because the paper said nothing about ethics anywhere; counting those as “did not engage the press” would conflate said nothing with said no.

Regex, fixed before the results were looked at, applied to ethics.disclosureDetail ‖ ethics.harmMitigation ‖ ethics.evidence.quote:

/\b(press|media|journalist|reporter|news outlet|newspaper|embargo|public relations|press release|publicity)\b/i

34 of 4,472 papers match (0.8%). Fold, first rule wins, residue printed in full:

The residue in full, both entries read by hand:

• PETS/2017/to-permit-or-not-to-permit-that-is-the-usability-question-crowdsourcing-mobile-a — “Used test credentials, arbitrary locations, non-identifiable media, and made contact-upload tasks optional.”
• NDSS/2026/flippyram-a-large-scale-study-of-rowhammer-prevalence — “Bootable media left user disks untouched”.

Both belong to the third family — the word “media” in a non-press sense — and the fold's regex simply does not enumerate those two phrasings. The rule was not widened to absorb them, so that the residue stays visible as a residue rather than being tuned to zero.

The two papers in the “press contact” family were read in full and neither is a press-engagement case:

• IMC/2016/recommended-for-you-a-first-look-at-content-recommendation-networks — “We reached out to the five companies examined in this study via their public press contacts, but … none responded.” A notification channel, not a press strategy.
• USENIX/2023/access-denied-assessing-physical-risks-to-internet-access-networks — anonymises networks or locations “when those details appear in the public press”. A redaction rule.

Hence the content page's claim: zero of 4,472.

bash scripts/pr_fulltext_grep.sh — two single passes over 20 GB of paper.cols.txt. –binary-files=text is mandatory: some .cols files are detected as binary by grep and are silently skipped otherwise, which under-counts the corpus (see Corpus).

Pass 1, 20 terms, papers matching any: 674 of 5,869 (11.5%). Per-term paper counts are in the script's committed output. The three largest are journalist (260), news outlet (81) and press release (76) — and reading them shows why the raw count is not the finding: press release is almost always a cited source (an FTC, ICANN or vendor press release in the references), and journalist is almost always a study population (journalists as at-risk users, e.g. 2015/USENIX/investigating-the-computer-security-practices-and-needs-of-journalists).

Pass 2, a deliberately broad self-reference regex over the same 5,869 files: 80 candidate papers, every one read by hand. Real hits: 3.

• 2026/USENIX/bridges-to-self-silent-web-to-app-tracking-on-mobile-via-localhost — [1Vlummens, Tim; Girish, Aniketh; Weerasekara, Nipuna; Zuiderveen Borgesius, Frederik; Acar, Gunes; Vallina-Rodriguez, Narseo (2026): "Bridges to Self: Silent Web-to-App Tracking on Mobile via Localhost", in: Proceedings of the USENIX Security Symposium. (Link)]
• 2024/IEEE-SP/sok-safer-digital-safety-research-involving-at-risk-users — [2Bellini, Rosanna; Tseng, Emily; Warford, Noel; Daffalla, Alaa; Matthews, Tara; Consolvo, Sunny; Woelfer, Jill Palzkill; Kelley, Patrick Gage; Mazurek, Michelle L.; Cuomo, Dana; Dell, Nicola; Ristenpart, Thomas (2024): "SoK: Safer Digital-Safety Research Involving At-Risk Users", in: Proceedings of the IEEE Symposium on Security and Privacy. (DOI)]
• 2023/PETS/usability-and-enforceability-of-global-privacy-control — [3Zimmeck, Sebastian; Wang, Oliver; Alicki, Kuba; Wang, Jocelyn; Eng, Sophie (2023): "Usability and Enforceability of Global Privacy Control", Proceedings on Privacy Enhancing Technologies 2023(2). (DOI)]

The other 57 are noise of four kinds, and the noise is instructive if you re-run this: our median matching our media without a word boundary; Mediatek; reporters meaning users who file abuse reports; and press releases as cited sources. The regex was left broad rather than tightened, because a false negative here would have cost the page its only two cases.

All 80 candidates, so that “every hit was read by hand” is checkable rather than asserted. The three real hits are marked.

corpus: 5859 extraction records, 7 venues (CCS, IMC, NDSS, PETS,
USENIX Security, TheWebConf, IEEE S&P), 2010-2026.

==========================================================================
A. Press and media in what papers say about their own conduct
==========================================================================
Population                                    N     Papers whose ethics text mentions press/media  Share
--------------------------------------------  ----  ---------------------------------------------  -----
empirical                                     5118  —                                              —
empirical ∧ has an ethics record              4472  34                                             0.8%
… of those, gave free-text disclosure detail  2870  32                                             1.1%
… of those, named a harm-mitigation step      3186  30                                             0.9%

What the press/media match actually is                                       Papers  Share of 34
---------------------------------------------------------------------------  ------  -----------
press contact with journalists or press offices                              2       5.9%
vendor disclosure embargo (not press)                                        16      47.1%
"social media"/"media payload"/"media documents" -- the word, not the press  14      41.2%
UNMAPPED RESIDUE                                                             2       5.9%

Every matching paper, printed in full -- the population is small enough that a
ranking would be dishonest and a list is the only fair summary:

  [vendor disclosure embargo (not press)]
    IEEE-SP/2019/spectre-attacks-exploiting-speculative-execution
      ...provided preliminary results to CPU vendors and affected companies and participated in an embargo. ‖ Coordinated responsible disclosure and embargo with industry. ‖ Using the practice of responsible disclosure, disjoint ...
    USENIX/2020/medusa-microarchitectural-data-leakage-via-automated-attack-synthesis
      ...Initial finding disclosed to Intel; embargo lasted until November 12, 2019. ‖ Responsible disclosure and publication embargo. ‖ We disclosed our initial finding to In...
    IEEE-SP/2021/crosstalk-speculative-data-leaks-across-cores-are-real
      ...Reported staging-buffer and RDRAND/RDSEED leakage to Intel; embargo followed. ‖ Coordinated vulnerability disclosure and embargo with Intel. ‖ We disclosed an initial PoC of staging buffer l...
    CCS/2023/ileakage-browser-based-timerless-speculative-execution-attacks-on-apple-devices
      ...ple and discussed countermeasures with Safari and Product Security teams. ‖ Maintained an embargo and coordinated countermeasures with Apple. ‖ We initially disclosed our findings to Apple on September 12, 2022. Apple ha...
    USENIX/2023/spectrem-exploiting-electromagnetic-emanations-during-transient-execution
      ...Disclosed findings to Arm; Arm acknowledged them and requested no embargo. ‖ We disclosed our findings to Arm on January 16th, 2023. Arm acknowledged these attacks, but did not request an embargo....
    USENIX/2023/bunnyhop-exploiting-the-instruction-prefetcher
      ...Reported results to Intel; Intel stated no embargo was required. ‖ The results in this paper were reported to Intel. The company responded that the issue is covered by its s...
    USENIX/2023/downfall-exploiting-speculative-data-gathering
      ...Reported findings to Intel; embargoed until August 2023 and coordinated public release. ‖ Coordinated responsible disclosure with Intel under embargo. ‖ We re...
    IEEE-SP/2024/deeptheft-stealing-dnn-model-architectures-through-power-side-channel
      ...Disclosed RAPL container exposure to Docker security team; team acknowledged and embargoed a fix. ‖ Reported the issue to Docker security team before publication. ‖ we have disclosed this security issue to the d...
    NDSS/2024/faults-in-our-bus-novel-bus-fault-attack-to-break-arm-trustzone
      ...OP-TEE security team and Linaro were notified; a 90-day embargo preceded public fixes. ‖ Coordinated disclosure, embargo, fix development, and testing. ‖ We notified the OP-TEE security ...
    USENIX/2024/security-and-privacy-analysis-of-samsungs-crowd-sourced-bluetooth-location-track
      ...Reported all issues to Samsung with a minimum 90-day embargo before publication. ‖ Coordinated disclosure and minimum 90-day embargo before publication. ‖ We have reported all the iss...
    IEEE-SP/2025/slap-data-speculation-attacks-via-load-address-prediction-on-apple-silicon
      ...and proof-of-concept code disclosed to Apple on May 24, 2024; Apple requested an extended embargo. ‖ Extended embargo requested by Apple after responsible disclosure. ‖ We disclosed our results to Apple on May 24, 2024. ...
    NDSS/2026/exploiting-tlbs-in-virtualized-gpus-for-cross-vm-side-channel-attacks
      ...Disclosed findings and proof-of-concept to NVIDIA; embargo lifted March 21, 2025. ‖ Responsible disclosure to NVIDIA and shared proof-of-concept. ‖ We have disclosed our findings to...
    PETS/2026/privacy-in-theory-bugs-in-practice-grey-box-auditing-of-differential-privacy-lib
      ...Reported vulnerabilities to maintainers under a 60-day disclosure embargo. ‖ Responsible disclosure and 60-day embargo for patches. ‖ We followed standard responsible disclosure protocols by repor...
    USENIX/2026/security-and-privacy-analysis-of-tiles-location-tracking-protocol
      ...Disclosed findings to Tile, observed a 90-day embargo, then extended it by 60 days. ‖ Sent no unnatural traffic and attacked only their own devices and infrastructure. ‖ we sen...
    USENIX/2026/lost-in-blockchain-address-misuse-hidden-cross-platform-risks-and-their-security
      ...d only public sources; omitted private keys and exploit scripts and planned a publication embargo. ‖ No human subjects were involved. We also plan to collaborate with these repository platforms to help them detect and wa...
    CCS/2025/heracles-chosen-plaintext-attack-on-amd-sev-snp
      ... the vulnerability and coordinated disclosure. ‖ Coordinated disclosure with AMD under an embargo. ‖ We reported our findings to AMD in January 2025. AMD acknowledged the vulnerability and asked for a coordinated disclos...
  ["social media"/"media payload"/"media documents" -- the word, not the press]
    USENIX/2018/analysis-of-privacy-protections-in-fitness-tracking-social-networks-or-you-can-r
      ...ave consulted our Institutional Review Board (IRB) to confirm that our analysis of social media posts does not meet the definition of human subjects research and thus does not require IRB approval....
    PETS/2019/because-i-was-told-so-much-linguistic-indicators-of-mental-health-status-on-twit
      ...cy and ethics in assessing a person's mental health status using their language on social media....
    USENIX/2020/what-twitter-knows-characterizing-ad-targeting-practices-user-perceptions-and-ad
      ...Recruitment notice explained the request to upload social-media data. ‖ Collected only three advertising files, omitted personally identifiable information, and redacted identifiers. ‖ Thi...
    IMC/2022/enabling-passive-measurement-of-zoom-performance-in-production-networks
      ...Network operator sanitized traces; one-way hashing and media-payload removal protected privacy. ‖ Anonymized with one-way hashing; operator removed personal data and media payloads. ‖ A...
    IMC/2023/flocking-to-mastodon-tracking-the-great-twitter-migration
      ...Collected exclusively publicly available data following established social-media research ethics. ‖ We have obtained a waiver from the ethics committee at the author's institution....
    USENIX/2023/unique-identification-of-50-000-virtual-reality-users-from-head-hand-motion-data
      ...BeatLeader notified users through official social media channels after initial results. ‖ Did not release the raw dataset; released de-identified and normalized data. ‖ the protoco...
    PETS/2024/onion-services-in-the-wild-a-study-of-deanonymization-attacks
      ...Used publicly accessible PACER and media documents; excluded media sentencing memorandums and omitted their URLs. ‖ Focused on investigative-method counts and did no...
    WWW/2024/bridging-or-breaking-impact-of-intergroup-interactions-on-religious-polarization
      ...Discussed risks of inferring religion from social-media names. ‖ Dropped verified users and non-personal names; discussed risks of identity inference. ‖ It is worth discussing the ...
    WWW/2024/mentallama-interpretable-mental-health-analysis-on-social-media-with-large-langu
      ...Public social-media data; texts anonymized, paraphrased, and obfuscated. ‖ Applied privacy protocols, anonymization, paraphrasing, and obfuscati...
    WWW/2024/characterizing-ethereum-upgradable-smart-contracts-and-their-security-implicatio
      ...Sent email alerts to contract creators and searched social media to report issues. ‖ Reported vulnerabilities to contract creators via EthMail and personal social media. ‖ We have also manu...
    WWW/2025/pirates-of-charity-exploring-donation-based-abuses-in-social-media-platforms
      ... any human subjects, including scammers. We collected publicly available data from social media profiles using API queries....
    PETS/2026/precarious-but-active-a-look-at-privacy-behaviors-in-chinese-transformative-fand
      ...Participants consented; public social-media users were not contacted and posts were paraphrased to prevent identification. ‖ Required overseas interviews, minimized ide...
    IMC/2025/exploration-of-the-dynamics-of-buy-and-sale-of-social-media-accounts
      ...Disclosed relevant profiles and information to all five social-media platforms in August 2024. ‖ Used only public data; avoided purchases, interactions, CAPTCHA bypasses, paywalls, and automati...
    WWW/2026/moral-outrage-shapes-commitments-beyond-attention-multimodal-moral-emotions-on-y
      ...This study examined how moral emotions in multimodal media content shape user engagement across three behavioral levels...
  [press contact with journalists or press offices]
    IMC/2016/recommended-for-you-a-first-look-at-content-recommendation-networks
      ...Researchers contacted the five examined companies through public press contacts; none responded. ‖ We reached out to the five companies examined in this study via their public press contacts, but...
    USENIX/2023/access-denied-assessing-physical-risks-to-internet-access-networks
      ...nymize networks or locations in case studies (§5) when those details appear in the public press....
  [UNMAPPED RESIDUE]
    PETS/2017/to-permit-or-not-to-permit-that-is-the-usability-question-crowdsourcing-mobile-a
      ...bmit identifiable content. ‖ Used test credentials, arbitrary locations, non-identifiable media, and made contact-upload tasks optional. ‖ This study was approved by our institution's ethics board. To protect participant...
    NDSS/2026/flippyram-a-large-scale-study-of-rowhammer-prevalence
      ...nformed about risks, collected data, privacy implications, and upload consent. ‖ Bootable media left user disks untouched; results stayed local without explicit upload consent. ‖ None of the institutions of the authors h...

==========================================================================
B. The sentence that travels: percentages in detection evidence quotes
==========================================================================
Quantity                                           Value
-------------------------------------------------  -----
papers with >=1 detection tuple                    5655
detection tuples in the corpus                     27241
tuples whose evidence quote contains a percentage  11523
papers contributing at least one such quote        4316

Of the percentage-bearing quotes                                         Tuples  Share of 11523
-----------------------------------------------------------------------  ------  --------------------------
name what the percentage is a percentage OF, in the same sentence        6003    52.1%
do not                                                                   5520    47.9%
carry a date or month in the same sentence                               270     2.3%
… of the undenominated, are a performance figure (accuracy/precision/…)  1690    30.6% of the undenominated

By paper (a paper counts as denominated if ANY of its percentage quotes is):
Papers                                N     Share of 4316
------------------------------------  ----  -------------
at least one denominated quote        3070  71.1%
every percentage quote undenominated  1246  28.9%
at least one dated quote              214   5.0%

Over time (tuples, not papers -- the unit here is the sentence). The last two
columns decompose the trend: if the fall in denomination were purely an artefact
of the field publishing more model-performance figures, the non-performance
series would be flat. It is not.
Bucket      Quotes with a %  Denominated  Dated  Perf share  Non-perf denominated
----------  ---------------  -----------  -----  ----------  --------------------
2010–2013   994              63.0%        2.0%   12.9%       67.8% (n=866)
2014–2017   1674             62.1%        1.9%   16.5%       68.4% (n=1398)
2018–2021   3023             54.6%        2.8%   17.7%       61.7% (n=2489)
2022–2024   3615             47.2%        2.4%   21.6%       53.8% (n=2835)
2025–2026*  2217             44.2%        2.1%   21.9%       51.3% (n=1732)

==========================================================================
C. legal.foundViolations -- how often a paper asserts a law was broken
==========================================================================
(1) worst case wins -- one row per paper, matches practices:legal_enforcement
legal.foundViolations  Papers (of 402 that assessed a law)  Share
---------------------  -----------------------------------  -----
yes                    94                                   23.4%
partial                83                                   20.6%
no                     74                                   18.4%
not-assessed           151                                  37.6%
not-stated             0                                    0.0%
  sums to 402 of 402

(2) any-tuple -- a paper counts once per distinct value; sums above 402
legal.foundViolations  Papers with >=1 such tuple  Share of 402
---------------------  --------------------------  ------------
not-assessed           177                         44.0%
yes                    94                          23.4%
partial                93                          23.1%
no                     75                          18.7%
not-stated             2                           0.5%

Population                      N    Has >=1 tuple with foundViolations=yes  Share
------------------------------  ---  --------------------------------------  -----
assessed a law                  402  94                                      23.4%
assessed a law AND ran a crawl  131  43                                      32.8%

Population: 27,241 detection tuples across 5,655 papers. 11,523 tuples have an evidence.quote containing /\d+(\.\d+)?\s?%/, contributed by 4,316 papers.

Two rules, either sufficient for “denominated”:

DENOM_RE     = /\d+(?:\.\d+)?\s?%[^.]{0,60}?\bof\b\s+[a-z0-9"'(]/i
DENOM_PRE_RE = /\bof\s+(the\s+)?[\d,]{2,}\s+[a-z]+[^.]{0,60}?\d+(?:\.\d+)?\s?%/i

and two classifiers:

PERF_RE = /\b(accuracy|accurate|precision|recall|f1|f-score|f-measure|auc|roc|true[- ]positive|
           false[- ]positive|success rate|error rate|overhead|detection rate|hit rate)\b/i
DATE_RE = /\b(19|20)\d{2}\b|\b(january|…|december)\b/i

By paper, of the 4,316: 3,070 (71.1%) have at least one denominated quote, 1,246 (28.9%) have none, 214 (5.0%) have at least one dated quote.

Decomposition of the year trend, added after a first draft asserted without evidence that the fall was caused by the rising share of performance figures:

Residue in both directions. –examples samples every 500th match so the sample spans the corpus rather than its first page. Called undenominated (correctly, in every sampled case — these are rates and scores with no population attached):

CCS/2010/dialog-based-payload-aggregation-for-intrusion-detection
  "while reducing the amount of data to be analyzed to about 1 %."
USENIX/2023/silent-bugs-matter-a-study-of-compiler-introduced-security-bugs
  "Specifically, 21.6% can cause system hanging (DoS), 50.8% lead to crashes, and 19.2% lead to
   information leaks."
IEEE-SP/2016/users-really-do-plug-in-usb-drives-they-find
  "These two numbers allow us to bound the attack's success rate to between 45-98%."
NDSS/2026/firmagent-…
  "FirmAgent achieved a 91% precision and successfully confirmed 182 real vulnerabilities…"

Called denominated:

WWW/2012/understanding-and-combating-link-farming-in-the-twitter-social-network
  "Within a span of 3 days, 65 (i.e., 13% out of 500) of the users responded by following our
   account back."
NDSS/2020/melting-pot-of-origins-…
  "As a result, we found that 73.6% of fingerprints leaked visit time."
PETS/2024/over-fences-and-into-yards-…
  "Only 32% of respondents knew that commercial satellites existed."

The sampled residue contains no misclassification in either direction, but 12 of 11,523 is not a validation set, and the rule's known weakness is visible in the third undenominated example: “success rate” is a denominated concept in context (“of the drives we dropped”), and the sentence still does not say so. That is the property the page is claiming, so the rule counting it as undenominated is correct for this purpose and would be wrong for most others.

The script prints two tabulations of the same field, because they answer different questions and an earlier draft of the content page mixed them up.

(1) worst case wins — one row per paper, taking the strongest value it records. Sums to 402. This is what the content page publishes and what Legal enforcement publishes, so the two pages now agree.

(2) any-tuple — a paper counts once per distinct value it uses. Sums above 402.

Of the 131 papers that assessed a law and ran a crawl, 43 (32.8%) have at least one foundViolations == yes tuple.

Two caveats the content page states and this page restates because they are easy to lose: foundViolations is an enum and so may be published as a percentage, but “assessed a law” is itself an extraction judgement, and 402 papers is a small population for a five-way split.

Two scripts, both deterministic and both re-runnable against the current data/extract/run1. Neither carries a figure forward from an earlier corpus.

node scripts/report_public_relations.mjs             # the three query sections
node scripts/report_public_relations.mjs --wiki      # DokuWiki tables
node scripts/report_public_relations.mjs --examples   # + sampled residue, both directions
bash  scripts/pr_fulltext_grep.sh                    # the 20 GB full-text sweep, ~4 min

-- RESIDUE: quotes the rule called UNDENOMINATED (read these before trusting it) --
  CCS/2010/dialog-based-payload-aggregation-for-intrusion-detection
    "while reducing the amount of data to be analyzed to about 1 %."
  WWW/2015/secrets-lies-and-account-recovery-lessons-from-the-use-of-personal-knowledge-que
    "SMS and email-based account recovery have a significantly higher chance of success: 81% for SMS vs 75% for Email vs 61% (US/English) down to 44% (France/French) for secret questions."
  IMC/2018/dissecting-apples-meta-cdn-during-an-ios-update
    "We observe the maximum traffic from the Apple AS to peak at 211%, from Limelight at 438% and Akamai at 113% on Sep. 19."
  WWW/2019/who-watches-the-watchmen-exploring-complaints-on-the-web
    "We find that DMCA notices make up 98.6% (1.05B URLs) of the dataset and a similar share of domains (97.8%)."
  NDSS/2021/deceptive-deletions-for-protecting-withdrawn-posts-on-social-media-platforms
    "even when we consider only two decoy posts for each damaging deletion the adversarial performance (F-score) drops to 42%"
  PETS/2022/understanding-utility-and-privacy-of-demographic-data-in-education-technology-by
    "Across the folds, the model had an average prediction accuracy of 73.3% ... The model ... had an average prediction accuracy of 76.2%"
  USENIX/2023/silent-bugs-matter-a-study-of-compiler-introduced-security-bugs
    "Specifically, 21.6% can cause system hanging (DoS), 50.8% lead to crashes, and 19.2% lead to information leaks."
  PETS/2024/detorrent-an-adversarial-padding-only-traffic-analysis-defense
    "we find that defense performance remains high, reducing Tik-Tok to a 21.5% accuracy with an 88.8% bandwidth overhead."
  IEEE-SP/2016/users-really-do-plug-in-usb-drives-they-find
    "These two numbers allow us to bound the attack's success rate to between 45-98%."
  NDSS/2026/firmagent-leveraging-fuzzing-to-assist-llm-agents-with-iot-firmware-vulnerability-discovery
    "FirmAgent achieved a 91% precision and successfully confirmed 182 real vulnerabilities, comprising 45 command injection and 137 buffer overflow vulnerabilities."
  NDSS/2025/secret-spilling-drive-leaking-user-behavior-through-ssd-contention
    "Figure 14 shows the classification heat map of the SSD K with the best results of 97.0 %, the open-world dataset is correctly classified with 78 % accuracy."
  IEEE-SP/2023/uncovering-user-interactions-on-smartphones-via-contactless-wireless-charging-si
    "WISERS achieves a 100% overall success rate in the 40 end-to-end attacks within at most five attempts"

-- quotes the rule called DENOMINATED --
  IMC/2010/estimating-and-sampling-graphs-with-multidimensional-random-walks
    "For example, in the Flickr graph 21% of the users belong to one or more special interest groups [26]."
  WWW/2012/understanding-and-combating-link-farming-in-the-twitter-social-network
    "Within a span of 3 days, 65 (i.e., 13% out of 500) of the users responded by following our account back."
  PETS/2015/blocking-resistant-communication-through-domain-fronting
    "In both cases, about 38% of packets are empty (mostly ACKs), with many packets near the usual TCP Maximum Segment Size of 1460 bytes."
  WWW/2017/can-you-spot-the-fakes-on-the-limitations-of-user-feedback-in-online-social-netw
    "We also find that up to 3.8% of members accepting connection requests show skill in identifying real accounts."
  IEEE-SP/2019/synesthesia-detecting-screen-content-via-remote-acoustic-side-channels
    "The resulting accuracy is 99.0%. For comparison, this classifier had a 99.4% accuracy on a similar-sized validation set of traces of the first screen."
  NDSS/2020/melting-pot-of-origins-compromising-the-intermediary-web-services-that-rehost-websites
    "As a result, we found that 73.6% of fingerprints leaked visit time."
  USENIX/2021/how-great-is-the-great-firewall-measuring-chinas-dns-censorship
    "99% of poisoned responses hit our machine within 364ms ahead of the legitimate ones."
  USENIX/2022/qcsd-a-quic-client-side-website-fingerprinting-defence-framework
    "At both 1 ms and 5 ms granularities LCSS indicated that both the defended and simulated time series were transmitting similar amounts of bytes (within ε = 150) for over 85% of the intervals."
  USENIX/2023/dscope-a-cloud-native-internet-telescope
    "50% of random scanning IPs ... after just 22min (90% after 72 min), while 50% of scanning ASNs ... after just 5 min (90% after 17 min)."
  PETS/2024/over-fences-and-into-yards-privacy-threats-and-concerns-of-commercial-satellites
    "Only 32% of respondents knew that commercial satellites existed."
  PETS/2025/intractable-cookie-crumbs-unveiling-the-nexus-of-stateful-banner-interaction-and
    "Across all categories and runs, we observe a consistent trend where nearly 60% of cookies have an expiry exceeding 10 days"
  CCS/2025/asymmetry-vulnerability-and-physical-attacks-on-online-map-construction-for-auto
    "Under our attack, UGR rises to 44% for both flashlight and patch attacks-an absolute increase of 17%"

==========================================================================
C. legal.foundViolations -- how often a paper asserts a law was broken
==========================================================================
(1) worst case wins -- one row per paper, matches practices:legal_enforcement
legal.foundViolations  Papers (of 402 that assessed a law)  Share
---------------------  -----------------------------------  -----
yes                    94                                   23.4%
partial                83                                   20.6%
no                     74                                   18.4%
not-assessed           151                                  37.6%
not-stated             0                                    0.0%
  sums to 402 of 402

(2) any-tuple -- a paper counts once per distinct value; sums above 402
legal.foundViolations  Papers with >=1 such tuple  Share of 402
---------------------  --------------------------  ------------
not-assessed           177                         44.0%
yes                    94                          23.4%
partial                93                          23.1%
no                     75                          18.7%
not-stated             2                           0.5%

Population                      N    Has >=1 tuple with foundViolations=yes  Share
------------------------------  ---  --------------------------------------  -----
assessed a law                  402  94                                      23.4%
assessed a law AND ran a crawl  131  43                                      32.8%

The run above is reproduced verbatim by –examples before this block; only the sampled residue is repeated here.

corpus root: /workspace/publications_dataset/data/fulltext
paper.cols.txt files: 5869
papers matching any term: 674

-- papers per term (a paper can match several) --
press release               76
press coverage               4
media coverage              58
news coverage               26
journalist                 260
media attention             37
press attention              2
popular press               18
mainstream media            21
tech press                   1
science communication        7
public relations            26
sensational                 27
misreport                   37
media report                54
news report                 73
reported in the media        8
media outlet                51
news outlet                 81
embargo                     59

-- SELF-REFERENCE: papers describing press engagement with THEIR OWN results --
Regex is deliberately broad; every hit is meant to be read by hand.
2013/WWW/two-years-of-short-urls-internet-measurement-security-threats-and-countermeasure/paper.cols.txt:1064:track the user. This analysis, however, mainly deals with FP7-ICT-257007 funded by the 7th FP. The opinions eximplications inherent to shortening services design. Our pressed in this paper are those of the authors and do not
2013/IEEE-SP/the-crossfire-attack/paper.cols.txt:1556:[44] FBI National Press Office, "Over one million potential victims of botnet cyber crime,"
2014/USENIX/man-vs-machine-practical-adversarial-detection-of-malicious-crowdsourcing-worker/paper.cols.txt:63: 1 Our work received approval from our local IRB review board.
2014/USENIX/security-analysis-of-a-full-body-scanner/paper.cols.txt:1084:[51] TSA Press Office. TSA takes next steps to
2022/IMC/revealing-the-evolution-of-a-cloud-provider-through-its-network-weather-map/paper.cols.txt:624: their network in a dedicated website [2]. These events could give
2022/NDSS/auto-draft-200/paper.cols.txt:552:with Samsung, our MediaTek implementation in F IRM W IRE is
2022/PETS/on-dark-patterns-and-manipulation-of-website-publishers-by-cmps/paper.cols.txt:255:each available consent pop-up version (free, paid, 4 Monitoring and analysis of the network cometc. - see Table 2) of a studied CMP, we installed munications. To detect tracking cookies and other susone version pop-up per dedicated website hosted on picious behaviors, we rely on the Firefox web developer
2022/USENIX/caring-about-sharing-user-perceptions-of-multiparty-data-sharing/paper.cols.txt:249:the survey. Our study received ethics approval from our institution's office of research ethics (ORE). See survey at
2022/USENIX/understanding-and-improving-usability-of-data-dashboards-for-simplified-privacy/paper.cols.txt:661:and our median participant contributed a total of 837.5 data using it? in Survey 1. 78.8% of our participants responded
2022/USENIX/automated-side-channel-analysis-of-media-software-with-manifold-learning/paper.cols.txt:418:of our media inputs are all over 10K. We visualize encoding
2022/USENIX/helping-hands-measuring-the-impact-of-a-large-threat-intelligence-sharing-commun/paper.cols.txt:224:process development, advisory writing, and media outreach. fixed schedule, and described drawing on newly registered
2022/WWW/link-black-box-detection-of-cross-site-scripting-vulnerabilities-using-reinforce/paper.cols.txt:480:second action, one of the four media tags (i.e., img, video, audio,
2022/IEEE-SP/heapster-analyzing-the-security-of-dynamic-allocators-for-monolithic-firmware-im/paper.cols.txt:915:and 1 hour median. It is worth noting that the analysis of only 17 analysis results show that all tested HMLs variants are vulnerable
2025/NDSS/the-unusual-suspects-studying-reasons-for-lacking-updates-in-wordpress/paper.cols.txt:377: and verify information about our studies on dedicated websites. (see Appendices E and F for the code books of S1 and S2 ).
2025/PETS/sheeps-clothing-wolfish-intent-automated-detection-and-evaluation-of-problematic/paper.cols.txt:1037:7.2.2 Per Exchange Analysis. Findings from our mediation analysis
2025/PETS/what-are-they-gonna-do-with-my-data-privacy-expectations-concerns-and-behaviors/paper.cols.txt:1383: VR platform owners. Social media outreach [1, 104] regarding
2025/PETS/privacy-perceptions-and-behaviors-towards-targeted-advertising-on-social-media-a/paper.cols.txt:949:the interplay between country and culture in our mediation analysis.
2025/PETS/privacy-perceptions-and-behaviors-towards-targeted-advertising-on-social-media-a/paper.cols.txt:950:We also included religion in our mediation analysis, as religion and
2025/USENIX/a-mixed-methods-study-of-open-source-software-maintainers-on-vulnerability-manag/paper.cols.txt:1327:practices most, reporters who provide the best-rated (e.g.,
2025/USENIX/understanding-how-users-prepare-for-and-react-to-smartphone-theft/paper.cols.txt:1275:7 Ethics Considerations • Help Website: A dedicated website offering additional
2025/WWW/the-agenda-setting-function-of-social-media/paper.cols.txt:150: and coverage bias: which issues are covered by the news and how
2025/WWW/pirates-of-charity-exploring-donation-based-abuses-in-social-media-platforms/paper.cols.txt:134:summary, our work received several positive acknowledgments
2025/WWW/causal-modeling-of-climate-activism-on-reddit/paper.cols.txt:611:bottom-up dynamics of the first wave of protests, that was successful despite not being covered by media during its build-up stage on
2025/WWW/7-days-later-analyzing-phishing-site-lifespan-after-detected/paper.cols.txt:404:a 1.76-hour median lifespan but a 30.5-hour GSB detection time.
2025/IEEE-SP/study-club-labor-union-or-start-up-characterizing-teams-and-collaboration-in-the/paper.cols.txt:289: Both stages of our study received approval from the first
2025/IEEE-SP/learning-from-censored-experiences-social-media-discussions-around-censorship-ci/paper.cols.txt:817:14 Persian VPN names, (2) VPN-Dedicated websites: We
2025/IEEE-SP/restricting-the-link-effects-of-focused-attention-and-time-delay-on-phishing-war/paper.cols.txt:326:Our study received approval from the University of
2024/NDSS/50-shades-of-support-a-device-centric-analysis-of-android-security-updates/paper.cols.txt:412:Nexus devices in a dedicated website [30]. Firmware images
2024/NDSS/drainclog-detecting-rogue-accounts-with-illegally-obtained-nfts-using-classifiers-learned-on-graphs/paper.cols.txt:283:draining are being sold between $29 - $149 on dedicated websites [34] and the darkweb [20]. The accessibility of draining
2024/PETS/simply-tell-me-how-on-trustworthiness-and-technology-acceptance-of-attribute-bas/paper.cols.txt:1010:Our mediation analysis in Appendix C highlighted that there exists
2024/USENIX/investigating-moderation-challenges-to-combating-hate-and-harassment-the-case-of/paper.cols.txt:1709:reporters who abuse the report feature)? C2: What would be the point of having a reporting
2024/USENIX/what-keeps-people-secure-is-that-they-met-the-security-team-deconstructing-drive/paper.cols.txt:543:news: "if incidents are picked up by the media, then it is often
2024/WWW/global-news-synchrony-and-diversity-during-the-start-of-the-covid-19-pandemic/paper.cols.txt:574:sity. To measure the similarity of the events covered by news media
2024/IEEE-SP/casual-users-and-rational-choices-within-differential-privacy/paper.cols.txt:1126:implicitly, our study received a very high rating of 4.58 on
2024/IEEE-SP/sok-safer-digital-safety-research-involving-at-risk-users/paper.cols.txt:1116:they do not overpromise and underdeliver. While doing so and we always tell them to contact the communications staff
2024/IEEE-SP/sok-safer-digital-safety-research-involving-at-risk-users/paper.cols.txt:1182:we received inquiries from reporters seeking our expertise. Here, we discuss how these friction points are both
2024/IEEE-SP/c-frame-characterizing-and-measuring-in-the-wild-captcha-attacks/paper.cols.txt:1070:lawsuit settlement [55]. These settlements are typically detailed on a dedicated website, outlining the specifics of
2024/CCS/do-anything-now-characterizing-and-evaluating-in-the-wild-jailbreak-prompts-on-l/paper.cols.txt:418:is a dedicated website for collecting jailbreak prompts. Users on
2024/CCS/i-had-sort-of-a-sense-that-i-was-always-being-watched-since-i-was-examining-inte/paper.cols.txt:393:surveys, and interviews to ensure the ethical consideration of participants. Our study received expedited full approval for all three
2024/CCS/trust-because-you-cant-verify-privacy-and-security-hurdles-in-education-technolo/paper.cols.txt:246: While our study received exempt status by our Institutional
2024/CCS/rolling-in-the-shadows-analyzing-the-extraction-of-mev-across-layer-2-rollups/paper.cols.txt:663:the number of liquidations and arbitrages over the course of our median values while the maximum values are much higher than the
2023/IMC/does-it-spin-on-the-adoption-and-use-of-quics-spin-bit/paper.cols.txt:658:and (3) hosting a dedicated website on our measurement IPs. The
2023/PETS/usability-and-enforceability-of-global-privacy-control/paper.cols.txt:973:our institution's IRB staff, legal counsel, communications staff, university administrators, and faculty colleagues about the procedure
2023/PETS/senrev-measurement-of-personal-information-disclosure-in-online-health-communiti/paper.cols.txt:1409:their invaluable feedback and comments. Our work received support from National Science Foundation (NSF) grants, including NSF
2023/USENIX/work-from-home-and-covid-19-trajectories-of-endpoint-security-management-in-a-se/paper.cols.txt:123:in which it operates. Our study received IRB approval from
2023/USENIX/millions-of-people-are-watching-you-understanding-the-digital-safety-needs-and-p/paper.cols.txt:945: "[If] there were a forum for creators where they know journalists who have other avenues for audience engagement
2023/USENIX/examining-power-dynamics-and-user-privacy-in-smart-technology-use-among-jordania/paper.cols.txt:341: Author 1 (the primary researcher) and Author 2 (the princi-To ensure ethical considerations, our study received approval
2023/USENIX/instructions-unclear-undefined-behaviour-in-cellular-network-specifications/paper.cols.txt:1030: Undefined behaviour Mediatek MT6768
2023/USENIX/instructions-unclear-undefined-behaviour-in-cellular-network-specifications/paper.cols.txt:1548: Undefined behaviour Mediatek MT6768 Mediatek MT6853V HiSilicon Kirin 820 5G Samsung Exynos 990 Qualcomm SM8250
2023/USENIX/instructions-unclear-undefined-behaviour-in-cellular-network-specifications/paper.cols.txt:1564: Undefined behaviour Mediatek MT6768 Mediatek MT6853V HiSilicon Kirin 820 5G Samsung Exynos 990 Qualcomm SM8250
2023/USENIX/instructions-unclear-undefined-behaviour-in-cellular-network-specifications/paper.cols.txt:1598: Undefined behaviour Mediatek MT6768 Mediatek MT6853V HiSilicon Kirin 820 5G Samsung Exynos 990 Qualcomm SM8250
2023/WWW/propaganda-politica-pagada-exploring-u-s-political-facebook-ads-en-espanol/paper.cols.txt:384: to entry for civil society groups and journalists who are the inparty to be left-leaning, and the Republican party to be right-leaning.
2023/WWW/an-empirical-study-of-the-usage-of-checksums-for-web-downloads/paper.cols.txt:516:fle hosting websites (i.e., dedicated websites that host many diferent fles produced by third parties, e.g., https://download.cnet.com,
2015/USENIX/investigating-the-computer-security-practices-and-needs-of-journalists/paper.cols.txt:262:newsrooms and journalists who deal regularly with international (including non-Western) sources and stories of
2012/IEEE-SP/lastor-a-low-latency-as-aware-tor-client/paper.cols.txt:1110:that our median false negative rate is 11%. This compares 0
2017/IMC/ethical-issues-in-research-using-datasets-of-illicit-origin/paper.cols.txt:780:of the journalists who used ICIJ's systems and IRB approved
2017/PETS/why-privacy-is-all-but-forgotten/paper.cols.txt:1325:Our study received approval from the ethics board of
2017/USENIX/when-the-weakest-link-is-strong-secure-collaboration-in-the-case-of-the-panama-p/paper.cols.txt:130:journalists who did not have them.
2017/USENIX/when-the-weakest-link-is-strong-secure-collaboration-in-the-case-of-the-panama-p/paper.cols.txt:1103:Lessons for Journalistic Collaborations. A key factor in the Panama Papers' security success was the reputation ICIJ had built for exclusive, high-impact investigations. Journalists approached for the Panama Papers'
2017/USENIX/exploring-user-perceptions-of-discrimination-in-online-targeted-advertising/paper.cols.txt:1131:critical topic of online algorithmic discrimination. Our media ad spending. eMarketer (2016). https:
2026/NDSS/tickets-to-hide-an-inside-look-into-the-anti-abuse-ecosystem-through-internal-abuse-data/paper.cols.txt:1112:reporters who file significant amounts of reports. Most reports
2026/NDSS/tickets-to-hide-an-inside-look-into-the-anti-abuse-ecosystem-through-internal-abuse-data/paper.cols.txt:1344: such as CSAM hotlines, or reporters who can pressure the
2026/PETS/contextual-intent-activists-privacy-considerations-for-collaborative-technology/paper.cols.txt:841:PETS except for social media outreach and only Group 21 even
2026/USENIX/end-to-end-encrypted-collaborative-documents/paper.cols.txt:78:tigative journalists who use collaborative documents to work near-real-time collaboration.
2026/USENIX/bridges-to-self-silent-web-to-app-tracking-on-mobile-via-localhost/paper.cols.txt:996:localhost-based tracking. On mobile, several studies have exfor helping with media outreach, HTTP Archive Project for
2026/USENIX/bridges-to-self-silent-web-to-app-tracking-on-mobile-via-localhost/paper.cols.txt:1069:Public Disclosure. Because the tracking technique was actively abused, we pursued public disclosure in parallel with
2026/USENIX/bridges-to-self-silent-web-to-app-tracking-on-mobile-via-localhost/paper.cols.txt:1070:affected vendor coordination before submission of this paper. Public disclosures increase societal awareness, and motivate and expedite mitigations. The public disclosure occurred
2026/USENIX/bridges-to-self-silent-web-to-app-tracking-on-mobile-via-localhost/paper.cols.txt:1071:through a dedicated website (https://localmess.github
2026/USENIX/bridges-to-self-silent-web-to-app-tracking-on-mobile-via-localhost/paper.cols.txt:1072:.io) with independent journalistic reporting. As part of the
2026/USENIX/bridges-to-self-silent-web-to-app-tracking-on-mobile-via-localhost/paper.cols.txt:1073:same release process, reporters asked Meta and Yandex's statements the day before the publication, giving them a chance
2026/WWW/community-fact-checks-do-not-break-follower-loyalty/paper.cols.txt:233:community notes and their status histories on a dedicated website1 .
2019/IMC/econ-modeling-the-network-to-improve-application-performance/paper.cols.txt:149:than 60%, even under dynamic network conditions. Our median
2019/IMC/econ-modeling-the-network-to-improve-application-performance/paper.cols.txt:768:works. Our median throughput prediction error for Azure, Southeast,
2019/IMC/econ-modeling-the-network-to-improve-application-performance/paper.cols.txt:785:write and 6MB for read) in the Northeast network. Our median
2019/IMC/econ-modeling-the-network-to-improve-application-performance/paper.cols.txt:792:outperforms all seven alternative models. Our median throughput
2019/IMC/econ-modeling-the-network-to-improve-application-performance/paper.cols.txt:922:forms other models irrespective of the number of parallel connections; further, our median error numbers are much lower, by about
2019/NDSS/quantity-vs-quality-evaluating-user-interest-profiles-using-ad-preference-managers/paper.cols.txt:1163:BlueKai, the overlap of our median participant's interest profile
2019/NDSS/quantity-vs-quality-evaluating-user-interest-profiles-using-ad-preference-managers/paper.cols.txt:1194:interests for our median participant. This suggests that Google,
2019/NDSS/quantity-vs-quality-evaluating-user-interest-profiles-using-ad-preference-managers/paper.cols.txt:1486:across APMs. Our results echo anecdotal findings from investigative journalists who have found their profiles from data brokers to be woefully inaccurate [48], [10]. Further, participants
2019/NDSS/measuring-the-facebook-advertising-ecosystem/paper.cols.txt:1476: The most innovative newsroom in journalism. And reporters who still knock on doors. ** "Something that started decades ago
2019/USENIX/less-is-more-quantifying-the-security-benefits-of-debloating-web-applications/paper.cols.txt:370:USENIX AssociationActions covered by MediaWiki tutorials: MediaWiki provides
2019/USENIX/less-is-more-quantifying-the-security-benefits-of-debloating-web-applications/paper.cols.txt:1442:vulnerabilities might only be exploitable under specific configurations. For example the set of five file-upload-related vulnerabilities in our MediaWiki dataset (marked as gray in Table 9) require
2019/WWW/privacy-preserving-crowd-sourcing-of-web-searches-with-private-data-donor/paper.cols.txt:978:Our study received ethics approval from University College London. leaves plenty of room to grow to more users.
2019/WWW/who-watches-the-watchmen-exploring-complaints-on-the-web/paper.cols.txt:378:Remove Your Media 7.28 0.92 2.45 346 bu
2019/WWW/who-watches-the-watchmen-exploring-complaints-on-the-web/paper.cols.txt:486: remove your media
2019/WWW/who-watches-the-watchmen-exploring-complaints-on-the-web/paper.cols.txt:611:mangapark.me 3 4 business 18 Remove Your Media (99.1%) 28 3,901
2019/WWW/who-watches-the-watchmen-exploring-complaints-on-the-web/paper.cols.txt:728: Remove your media
2019/WWW/who-watches-the-watchmen-exploring-complaints-on-the-web/paper.cols.txt:729: Remove your media
2021/IMC/identifying-ases-of-state-owned-internet-operators/paper.cols.txt:803: a dedicated website.
2021/NDSS/pfirewall-semantics-aware-customizable-data-flow-control-for-smart-home-privacy-protection/paper.cols.txt:1013:that our mediator works correctly in relaying IoT device events
2021/USENIX/the-role-of-computer-security-customer-support-in-helping-survivors-of-intimate/paper.cols.txt:253:Ethical Considerations. Our study received IRB approval.
2021/WWW/an-experimental-study-to-understand-user-experience-and-perception-bias-occurred/paper.cols.txt:466:the pre-stance and the four media type combinations, this figure quantifying the difference between two pre-stances, positive and
2021/WWW/short-is-the-road-that-leads-from-fear-to-hate-fear-speech-in-indian-whatsapp-gr/paper.cols.txt:363:Tyson [21] to gather the WhatsApp data. With help from journalists who cover politics, we curated lists of keywords related to
2021/WWW/security-of-alerting-authorities-in-the-www-measuring-namespaces-dnssec-and-web/paper.cols.txt:194:own dedicated websites not only for informational purposes but
2020/WWW/an-empirical-study-of-the-use-of-integrity-verification-mechanisms-for-web-subre/paper.cols.txt:1124:and to promote its use through a dedicated website that would
2018/PETS/panoptispy-characterizing-audio-and-video-exfiltration-from-android-applications/paper.cols.txt:934:ative rate of our media-detection methodology is low.
2018/PETS/panoptispy-characterizing-audio-and-video-exfiltration-from-android-applications/paper.cols.txt:1134:Dynamic analysis limitations. A number of factors could lead to this result. First, our media extraction
2018/USENIX/understanding-the-reproducibility-of-crowd-reported-security-vulnerabilities/paper.cols.txt:1249:through cash incentives. For public disclosure websites,
2018/USENIX/end-to-end-measurements-of-email-spoofing-attacks/paper.cols.txt:1085:Ethics Guidelines. Our study received IRB approval, compared to that without security indicators. The differand we have taken active steps to protect the participants. ence, however, is not very significant (Fisher's exact test
2018/USENIX/saql-a-stream-based-query-system-for-real-time-abnormal-system-behavior-detectio/paper.cols.txt:223:low alert detection latency (<2s); (2) our pressure test
2018/USENIX/all-your-gps-are-belong-to-us-towards-stealthy-manipulation-of-road-navigation-s/paper.cols.txt:119: 1 Our study received the approval from our local IRB (#17-936).
2018/USENIX/all-your-gps-are-belong-to-us-towards-stealthy-manipulation-of-road-navigation-s/paper.cols.txt:870:customized driving simulator. Our study received the approval of our local IRB (#17-936).
2018/IEEE-SP/hackers-vs-testers-a-comparison-of-software-vulnerability-discovery-processes/paper.cols.txt:271:To identify reporters who successfully submitted vulnerabilities,
2018/CCS/minesweeper-an-in-depth-look-into-drive-by-cryptocurrency-mining-and-its-defense/paper.cols.txt:83:such as UNICEF [40], launched dedicated websites to mine for donations, and legitimate websites are exploring mining in an attempt to
2016/PETS/individual-versus-organizational-computer-security-and-privacy-concerns-in-journ/paper.cols.txt:423:even journalists who understand the risks of insecure
2016/PETS/individual-versus-organizational-computer-security-and-privacy-concerns-in-journ/paper.cols.txt:973:replicated this study with journalists who live in countries with weaker press and speech protections. However, because of Internet technology and the globalization of the media (e.g., [46]) we expect that some of our
papers with a self-reference hit: 80
(end of self-reference hits)

2012/IEEE-SP/lastor-a-low-latency-as-aware-tor-client
2013/IEEE-SP/the-crossfire-attack
2013/WWW/two-years-of-short-urls-internet-measurement-security-threats-and-countermeasure
2014/USENIX/man-vs-machine-practical-adversarial-detection-of-malicious-crowdsourcing-worker
2014/USENIX/security-analysis-of-a-full-body-scanner
2015/USENIX/investigating-the-computer-security-practices-and-needs-of-journalists
2016/PETS/individual-versus-organizational-computer-security-and-privacy-concerns-in-journ
2017/IMC/ethical-issues-in-research-using-datasets-of-illicit-origin
2017/PETS/why-privacy-is-all-but-forgotten
2017/USENIX/exploring-user-perceptions-of-discrimination-in-online-targeted-advertising
2017/USENIX/when-the-weakest-link-is-strong-secure-collaboration-in-the-case-of-the-panama-p
2018/CCS/minesweeper-an-in-depth-look-into-drive-by-cryptocurrency-mining-and-its-defense
2018/IEEE-SP/hackers-vs-testers-a-comparison-of-software-vulnerability-discovery-processes
2018/PETS/panoptispy-characterizing-audio-and-video-exfiltration-from-android-applications
2018/USENIX/all-your-gps-are-belong-to-us-towards-stealthy-manipulation-of-road-navigation-s
2018/USENIX/end-to-end-measurements-of-email-spoofing-attacks
2018/USENIX/saql-a-stream-based-query-system-for-real-time-abnormal-system-behavior-detectio
2018/USENIX/understanding-the-reproducibility-of-crowd-reported-security-vulnerabilities
2019/IMC/econ-modeling-the-network-to-improve-application-performance
2019/NDSS/measuring-the-facebook-advertising-ecosystem
2019/NDSS/quantity-vs-quality-evaluating-user-interest-profiles-using-ad-preference-managers
2019/USENIX/less-is-more-quantifying-the-security-benefits-of-debloating-web-applications
2019/WWW/privacy-preserving-crowd-sourcing-of-web-searches-with-private-data-donor
2019/WWW/who-watches-the-watchmen-exploring-complaints-on-the-web
2020/WWW/an-empirical-study-of-the-use-of-integrity-verification-mechanisms-for-web-subre
2021/IMC/identifying-ases-of-state-owned-internet-operators
2021/NDSS/pfirewall-semantics-aware-customizable-data-flow-control-for-smart-home-privacy-protection
2021/USENIX/the-role-of-computer-security-customer-support-in-helping-survivors-of-intimate
2021/WWW/an-experimental-study-to-understand-user-experience-and-perception-bias-occurred
2021/WWW/security-of-alerting-authorities-in-the-www-measuring-namespaces-dnssec-and-web
2021/WWW/short-is-the-road-that-leads-from-fear-to-hate-fear-speech-in-indian-whatsapp-gr
2022/IEEE-SP/heapster-analyzing-the-security-of-dynamic-allocators-for-monolithic-firmware-im
2022/IMC/revealing-the-evolution-of-a-cloud-provider-through-its-network-weather-map
2022/NDSS/auto-draft-200
2022/PETS/on-dark-patterns-and-manipulation-of-website-publishers-by-cmps
2022/USENIX/automated-side-channel-analysis-of-media-software-with-manifold-learning
2022/USENIX/caring-about-sharing-user-perceptions-of-multiparty-data-sharing
2022/USENIX/helping-hands-measuring-the-impact-of-a-large-threat-intelligence-sharing-commun
2022/USENIX/understanding-and-improving-usability-of-data-dashboards-for-simplified-privacy
2022/WWW/link-black-box-detection-of-cross-site-scripting-vulnerabilities-using-reinforce
2023/IMC/does-it-spin-on-the-adoption-and-use-of-quics-spin-bit
2023/PETS/senrev-measurement-of-personal-information-disclosure-in-online-health-communiti
2023/PETS/usability-and-enforceability-of-global-privacy-control   <-- REAL
2023/USENIX/examining-power-dynamics-and-user-privacy-in-smart-technology-use-among-jordania
2023/USENIX/instructions-unclear-undefined-behaviour-in-cellular-network-specifications
2023/USENIX/millions-of-people-are-watching-you-understanding-the-digital-safety-needs-and-p
2023/USENIX/work-from-home-and-covid-19-trajectories-of-endpoint-security-management-in-a-se
2023/WWW/an-empirical-study-of-the-usage-of-checksums-for-web-downloads
2023/WWW/propaganda-politica-pagada-exploring-u-s-political-facebook-ads-en-espanol
2024/CCS/do-anything-now-characterizing-and-evaluating-in-the-wild-jailbreak-prompts-on-l
2024/CCS/i-had-sort-of-a-sense-that-i-was-always-being-watched-since-i-was-examining-inte
2024/CCS/rolling-in-the-shadows-analyzing-the-extraction-of-mev-across-layer-2-rollups
2024/CCS/trust-because-you-cant-verify-privacy-and-security-hurdles-in-education-technolo
2024/IEEE-SP/c-frame-characterizing-and-measuring-in-the-wild-captcha-attacks
2024/IEEE-SP/casual-users-and-rational-choices-within-differential-privacy
2024/IEEE-SP/sok-safer-digital-safety-research-involving-at-risk-users   <-- REAL
2024/NDSS/50-shades-of-support-a-device-centric-analysis-of-android-security-updates
2024/NDSS/drainclog-detecting-rogue-accounts-with-illegally-obtained-nfts-using-classifiers-learned-on-graphs
2024/PETS/simply-tell-me-how-on-trustworthiness-and-technology-acceptance-of-attribute-bas
2024/USENIX/investigating-moderation-challenges-to-combating-hate-and-harassment-the-case-of
2024/USENIX/what-keeps-people-secure-is-that-they-met-the-security-team-deconstructing-drive
2024/WWW/global-news-synchrony-and-diversity-during-the-start-of-the-covid-19-pandemic
2025/IEEE-SP/learning-from-censored-experiences-social-media-discussions-around-censorship-ci
2025/IEEE-SP/restricting-the-link-effects-of-focused-attention-and-time-delay-on-phishing-war
2025/IEEE-SP/study-club-labor-union-or-start-up-characterizing-teams-and-collaboration-in-the
2025/NDSS/the-unusual-suspects-studying-reasons-for-lacking-updates-in-wordpress
2025/PETS/privacy-perceptions-and-behaviors-towards-targeted-advertising-on-social-media-a
2025/PETS/sheeps-clothing-wolfish-intent-automated-detection-and-evaluation-of-problematic
2025/PETS/what-are-they-gonna-do-with-my-data-privacy-expectations-concerns-and-behaviors
2025/USENIX/a-mixed-methods-study-of-open-source-software-maintainers-on-vulnerability-manag
2025/USENIX/understanding-how-users-prepare-for-and-react-to-smartphone-theft
2025/WWW/7-days-later-analyzing-phishing-site-lifespan-after-detected
2025/WWW/causal-modeling-of-climate-activism-on-reddit
2025/WWW/pirates-of-charity-exploring-donation-based-abuses-in-social-media-platforms
2025/WWW/the-agenda-setting-function-of-social-media
2026/NDSS/tickets-to-hide-an-inside-look-into-the-anti-abuse-ecosystem-through-internal-abuse-data
2026/PETS/contextual-intent-activists-privacy-considerations-for-collaborative-technology
2026/USENIX/bridges-to-self-silent-web-to-app-tracking-on-mobile-via-localhost   <-- REAL
2026/USENIX/end-to-end-encrypted-collaborative-documents
2026/WWW/community-fact-checks-do-not-break-follower-loyalty

Every quote on the content page was checked against the source. Two needed the PDF text layer rather than paper.cols.txt:

Generated with node scripts/bibgen.mjs from data/corpus2/.meta (publisher metadata, real DOIs), not from recall. Two USENIX records carry no authors in the index — 100% of USENIX and PETS records do not — so authors were fetched from the USENIX landing page with curl and a browser User-Agent (WebFetch gets 403 from usenix.org), read out of the citation_author meta tags.

zimmeck2023_gpc and gamerogarrido2017_legalrisks were already in Bibliography and were reused unchanged. 11 keys were added; all were checked against the 351 existing entries for collisions before appending, and none collided.

Two Sonnet sub-agents did the fetching, both instructed to use primary sources only, to report a fetch date for every claim, and to return a rejected list. Every load-bearing quote below was then re-fetched and re-verified by the orchestrating model before it went on the page; two are noted as verified-by-agent-only.

Method note worth carrying to the next run: WebFetch returned wrong article content for several PMC and BMJ URLs during this run (unrelated crystallography and cardiology papers), and hit Cloudflare or 403 on bmj.com, acm.org, pubmed.ncbi.nlm.nih.gov and usenix.org. Everything load-bearing was therefore cross-checked through a second channel — NCBI E-utilities, the Crossref API, or Europe PMC full-text XML, via curl. Do not trust a single WebFetch response for a numbered claim.

Four reviewers, all told explicitly that the orchestrator's context might not be exhaustive, all handed the page text, both scripts, their unedited output and these notes. The three focused ones ran in parallel first; the generic one ran afterwards on the corrected text.

Re-ran both scripts and diffed against the committed output: byte-identical. Checked every figure on both pages.

The most productive of the four. It was given no checklist and told the other three had already covered figures, citations and external currency.

The figures reviewer and the citations reviewer were both re-run against the revised text, since findings from both had been acted on. Both came back clean.

• Figures, re-run. Both scripts re-run from scratch and diffed byte-for-byte against the committed output; the worst case wins precedence checked and confirmed to sum to 402 and to agree with Legal enforcement; the 80-candidate sweep reproduced and diffed slug-for-slug against the list embedded on this page; the 20-term count, all five year-bucket rows including the two new columns, and the [11Saemann, Marlene; Theis, Daniel; Urban, Tobias; Degeling, Martin (2022): "Investigating GDPR Fines in the Light of Data Flows", Proceedings on Privacy Enhancing Technologies 2022(4). (DOI)] and [1Vlummens, Tim; Girish, Aniketh; Weerasekara, Nipuna; Zuiderveen Borgesius, Frederik; Acar, Gunes; Vallina-Rodriguez, Narseo (2026): "Bridges to Self: Silent Web-to-App Tracking on Mobile via Localhost", in: Proceedings of the USENIX Security Symposium. (Link)] figures all verified against source. No blocking or should-fix findings. It independently re-derived the saemann2022_fines 45%-vs-37.6% discrepancy and confirmed the footnote already discloses it.
• Citations, re-run. bellini2024_safer confirmed correct in both shipped artefacts this time — the .bib and the full spliced bibliography — against Crossref. The bibliography diff confirmed as a pure append: 351 → 362 entries, block inserted before </bibtex>, no other line changed. All 15 keys resolve exactly once. All quotes re-verified verbatim, including the two-column-split ones. No findings. It noted that the saemann2022_fines paragraph was being edited while it read, and re-read to a stable state before reporting — worth knowing if you re-run reviewers against a live workdir.

What this reviewer's findings say about the other three. All three focused reviewers passed the page while its headline claim was false, because none of them was asked whether the claim was true — only whether it matched the script, the sources and the live web. It did, on all three counts: the script faithfully reported what the regex found, and the regex was wrong. The generic slot earned its place in this run and should not be dropped.

• Whether press engagement is actually rare, or just unreported. The corpus can only show the second. CHI and SOUPS are absent from the corpus and are where the at-risk and usable-security communities publish — the one paper here with press advice is from that community. Re-running section A over CHI and SOUPS is the single highest-value follow-up.
• Whether a paper's own publicity contaminates its follow-up measurement of the same population. Nothing in the corpus measures this. It is flagged on the content page as an open question with a footnote, not as a finding.
• Any figure at all on how often a measurement result is misreported. No paper in these seven venues measures the accuracy of press coverage of security or privacy research.
• The true number of papers that document their own press handling. Three is a floor from a keyword sweep that is now known to have under-recalled once. The only way to raise confidence is a different method — e.g. classifying every ethics and Disclosure section rather than grepping — which was not run.
• Why the denomination rate is falling. The script now decomposes the trend against the obvious confounder — the rising share of model-performance figures — and the confounder does not account for it: performance figures rise from 12.9% to 21.9% of sentences, while denomination among non-performance sentences falls from 67.8% (n=866) to 51.3% (n=1,732). So the fall is not an artefact of subject matter alone. What it is remains unestablished; plausible candidates the corpus cannot separate are longer and more compressed result sentences, more tables carrying the denominator instead of the prose, and drift in what the extraction model selects as an evidence quote. The content page says only that the fall survives the decomposition.

Back to the content page: Public relations. Corpus-wide caveats: Corpus.

[1]

Vlummens, Tim; Girish, Aniketh; Weerasekara, Nipuna; Zuiderveen Borgesius, Frederik; Acar, Gunes; Vallina-Rodriguez, Narseo (2026): "Bridges to Self: Silent Web-to-App Tracking on Mobile via Localhost", in: Proceedings of the USENIX Security Symposium. (Link)

[2]

Bellini, Rosanna; Tseng, Emily; Warford, Noel; Daffalla, Alaa; Matthews, Tara; Consolvo, Sunny; Woelfer, Jill Palzkill; Kelley, Patrick Gage; Mazurek, Michelle L.; Cuomo, Dana; Dell, Nicola; Ristenpart, Thomas (2024): "SoK: Safer Digital-Safety Research Involving At-Risk Users", in: Proceedings of the IEEE Symposium on Security and Privacy. (DOI)

[3]

Zimmeck, Sebastian; Wang, Oliver; Alicki, Kuba; Wang, Jocelyn; Eng, Sophie (2023): "Usability and Enforceability of Global Privacy Control", Proceedings on Privacy Enhancing Technologies 2023(2). (DOI)

[4]

Perez, Daniel; Livshits, Benjamin (2021): "Smart Contract Vulnerabilities: Vulnerable Does Not Imply Exploited", in: Proceedings of the USENIX Security Symposium. (Link)

[5]

Kotzias, Platon; Razaghpanah, Abbas; Amann, Johanna; Paterson, Kenneth G.; Vallina-Rodriguez, Narseo; Caballero, Juan (2018): "Coming of Age: A Longitudinal Study of TLS Deployment", in: Proceedings of the ACM Internet Measurement Conference. (DOI)

[6]

Deuber, Dominic; Keuchen, Michael; Christin, Nicolas (2023): "Assessing Anonymity Techniques Employed in German Court Decisions: A De-Anonymization Experiment", in: Proceedings of the USENIX Security Symposium. (Link)

[7]

Pan, Elleen; Ren, Jingjing; Lindorfer, Martina; Wilson, Christo; Choffnes, David (2018): "Panoptispy: Characterizing Audio and Video Exfiltration from Android Applications", Proceedings on Privacy Enhancing Technologies 2018(4):33-50. (DOI)

[8]

Dubois, Daniel J.; Kolcun, Roman; Mandalari, Anna Maria; Paracha, Muhammad Talha; Choffnes, David; Haddadi, Hamed (2020): "When Speakers Are All Ears: Characterizing Misactivations of IoT Smart Speakers", Proceedings on Privacy Enhancing Technologies 2020(4):255-276. (DOI)

[9]

Cui, Ang; Costello, Michael; Stolfo, Salvatore J. (2013): "When Firmware Modifications Attack: A Case Study of Embedded Exploitation", in: Proceedings of the Network and Distributed System Security Symposium. (Link)

[10]

Smith, C. Estelle; Nevarez, Eduardo; Zhu, Haiyi (2020): "Disseminating Research News in HCI: Perceived Hazards, How-To's, and Opportunities for Innovation", in: Proceedings of the CHI Conference on Human Factors in Computing Systems. (DOI)

[11]

Saemann, Marlene; Theis, Daniel; Urban, Tobias; Degeling, Martin (2022): "Investigating GDPR Fines in the Light of Data Flows", Proceedings on Privacy Enhancing Technologies 2022(4). (DOI)

[12]

Sumner, Petroc; Vivian-Griffiths, Solveiga; Boivin, Jacky; Williams, Andy; Venetis, Christos A.; Davies, Aimée; Ogden, Jack; Whelan, Leanne; Hughes, Bethan; Dalton, Bethan; Boy, Fred; Chambers, Christopher D. (2014): "The Association Between Exaggeration in Health Related Science News and Academic Press Releases: Retrospective Observational Study", BMJ 349:g7015. (DOI)

[13]

Bratton, Luke; Adams, Rachel C.; Challenger, Aimée; Boivin, Jacky; Bott, Lewis; Chambers, Christopher D.; Sumner, Petroc (2019): "The Association Between Exaggeration in Health-Related Science News and Academic Press Releases: A Replication Study", Wellcome Open Research 4:148. (DOI)

[14]

Adams, Rachel C.; Challenger, Aimée; Bratton, Luke; Boivin, Jacky; Bott, Lewis; Powell, Georgina; Williams, Andy; Chambers, Christopher D.; Sumner, Petroc (2019): "Claims of Causality in Health News: A Randomised Trial", BMC Medicine 17:91. (DOI)