provenance:privacy:email_tracking
Differences
This shows you the differences between two versions of the page.
| Both sides previous revisionPrevious revisionNext revision | Previous revision | ||
| provenance:privacy:email_tracking [2026/09/02 07:41] – Add sections 14.6-14.8: the generic (Fable) pass, the re-verification pass, and the prose-mode attribution guard that came out of it. Fix the provenance page's own three-way disagreement about its figure count, the stale script annotations it republishes, karel.kubicek.claude | provenance:privacy:email_tracking [2026/09/21 14:02] (current) – Quote-check refresh 2026-09-21: re-ran report_email_tracking.mjs; 62 below threshold -> 49 rescued from the PDF + 13 below in both. Authored by Claude karel.kubicek.claude | ||
|---|---|---|---|
| Line 160: | Line 160: | ||
| ===== 8. Quote checks ===== | ===== 8. Quote checks ===== | ||
| - | Every '' | + | Every '' |
| ^ Verdict ^ Quotes ^ | ^ Verdict ^ Quotes ^ | ||
| | verbatim | 220 | | | verbatim | 220 | | ||
| | located after normalisation | 118 | | | located after normalisation | 118 | | ||
| - | | below threshold | 62 | | + | | rescued from the PDF | 49 | |
| + | | below threshold | ||
| | **total** | **400** | | | **total** | **400** | | ||
| - | **Below threshold | + | **The old figure was 62 below threshold, and 49 of those 62 are a defect |
| - | **Eight were read by hand** against '' | + | **Below threshold is not " |
| + | |||
| + | **Eight | ||
| ^ Paper ^ Coverage ^ Longest contiguous run found ^ Verdict ^ | ^ Paper ^ Coverage ^ Longest contiguous run found ^ Verdict ^ | ||
| Line 188: | Line 191: | ||
| | {[utz2023_comparing]} | 20% | //"No HTTPS was rarest, / with 2.85 % of sites"// | | {[utz2023_comparing]} | 20% | //"No HTTPS was rarest, / with 2.85 % of sites"// | ||
| - | Four of the 62 are below 25% — {[tu2016_security]} (23%), {[utz2023_comparing]} (20%), {[reaves2016_sending]} (20%) and {[agarwal2025_fishing]} (18%). Checking that claim is how '' | + | Four of the then-62 were below 25% — {[tu2016_security]} (23%), {[utz2023_comparing]} (20%), {[reaves2016_sending]} (20%) and {[agarwal2025_fishing]} (18%). Checking that claim is how '' |
| ===== 9. External sources, and how each was verified ===== | ===== 9. External sources, and how each was verified ===== | ||
| Line 1053: | Line 1056: | ||
| verbatim | verbatim | ||
| located after normalisation | located after normalisation | ||
| - | below threshold | + | |
| + | | ||
| Below threshold is NOT " | Below threshold is NOT " | ||
| - | USENIX/ | ||
| - | 53% of 5-word windows: Based on the total number of visitors where we have referrer information, | ||
| - | USENIX/ | ||
| - | 55% of 5-word windows: Through our analysis we discovered vulnerabilities in the feature phone platforms of all major manufacturers.… | ||
| - | IMC/ | ||
| - | 50% of 5-word windows: On average, spammers make many more phone calls than legitimate users, however the average number of phone cal… | ||
| USENIX/ | USENIX/ | ||
| 42% of 5-word windows: We find that indeed spam exposure leads to significantly, | 42% of 5-word windows: We find that indeed spam exposure leads to significantly, | ||
| - | USENIX/ | ||
| - | 56% of 5-word windows: repeat orders are an important part of the business, constituting 27% and 38% of average program revenue for G… | ||
| USENIX/ | USENIX/ | ||
| 54% of 5-word windows: The graph shows that just 10% of the highest-revenue affiliates account for 75-90% of total program revenue… | 54% of 5-word windows: The graph shows that just 10% of the highest-revenue affiliates account for 75-90% of total program revenue… | ||
| - | USENIX/ | ||
| - | 38% of 5-word windows: the median annualized affiliate commissions for GlavMed, SpamIt, and RX-Promotion are just $292, $3,320, and $… | ||
| - | NDSS/ | ||
| - | 56% of 5-word windows: There were 1,500 calls made on 30th March 2014 and another 1300 on 31st March. Incidentally, | ||
| - | NDSS/ | ||
| - | 56% of 5-word windows: the majority of the phone numbers, 91%, are registered within the United States… | ||
| - | CCS/ | ||
| - | 45% of 5-word windows: when the malware sends SIP messages to the SMS client with the speed at least 3,825 messages ... per second, t… | ||
| - | CCS/ | ||
| - | 50% of 5-word windows: We employ DonateARC to make an unauthorized $10 donation to ARC from the victim' | ||
| CCS/ | CCS/ | ||
| 23% of 5-word windows: The snapshot of the victim account' | 23% of 5-word windows: The snapshot of the victim account' | ||
| Line 1083: | Line 1069: | ||
| PETS/ | PETS/ | ||
| 43% of 5-word windows: Our crawler identified a contact page on 29.6% of the 100K websites... successfully located contact forms on 1… | 43% of 5-word windows: Our crawler identified a contact page on 29.6% of the 100K websites... successfully located contact forms on 1… | ||
| - | PETS/ | ||
| - | 54% of 5-word windows: we discovered that 6.1% (1,035) of all submitted contact forms leak a user's PII to third parties.… | ||
| - | PETS/ | ||
| - | 47% of 5-word windows: Over the duration of our experiment we received 309 emails from third parties, that is, domains which our craw… | ||
| - | USENIX/ | ||
| - | 56% of 5-word windows: out of 35,832 transmitted vulnerability reports, only 2,064 (5.8%) were actually received successfully… | ||
| - | USENIX/ | ||
| - | 57% of 5-word windows: If a notification report is read by the owner of the vulnerable application, | ||
| - | NDSS/ | ||
| - | 58% of 5-word windows: we conclude that Set-4 includes at least 4.1% of suspicious messages, corresponding to δ = 5.0.… | ||
| - | NDSS/ | ||
| - | 50% of 5-word windows: Set-5 includes at least 0.39% of suspicious messages, using the most conservative combination Vmin + threshold… | ||
| - | NDSS/ | ||
| - | 56% of 5-word windows: the median (mean) deviation distance of FBS localization is only 11 (149) meters.… | ||
| PETS/ | PETS/ | ||
| 40% of 5-word windows: The average Jaccard similarity between the sets of third parties loaded during the first and second views of t… | 40% of 5-word windows: The average Jaccard similarity between the sets of third parties loaded during the first and second views of t… | ||
| - | PETS/ | ||
| - | 50% of 5-word windows: We found that the previously-unused phone number became targetable in 36 days, showing that it had indeed been… | ||
| PETS/ | PETS/ | ||
| 52% of 5-word windows: We found seven loosely-integrated third-party scripts collecting Facebook user data using the first party' | 52% of 5-word windows: We found seven loosely-integrated third-party scripts collecting Facebook user data using the first party' | ||
| - | USENIX/ | ||
| - | 58% of 5-word windows: 56.6 % of notified owners remediated the issue, compared to 9.2 % in the control group.… | ||
| - | USENIX/ | ||
| - | 56% of 5-word windows: Under this assumption, half of the notified owners (46.9 %) have used the tool at least once for their site(s)… | ||
| - | CCS/ | ||
| - | 59% of 5-word windows: On randomly sampled 1000 SMS images, STR achieves an accuracy of 90%, i.e. accurately recognizing 90% of messa… | ||
| - | CCS/ | ||
| - | 45% of 5-word windows: Among these 280 spam campaigns, 53 of them contain SMS spam messages in different languages.… | ||
| - | CCS/ | ||
| - | 56% of 5-word windows: As a result, it falsely reported 103 out of 124 benign SMS messages.… | ||
| - | PETS/ | ||
| - | 55% of 5-word windows: 53% of the previously vulnerable organizations have not corrected their policy and an additional 27% of previo… | ||
| - | PETS/ | ||
| - | 50% of 5-word windows: We conclude that all data responses of the vulnerable organizations contained sensitive personal information… | ||
| WWW/ | WWW/ | ||
| 50% of 5-word windows: After the five S→I interactions, | 50% of 5-word windows: After the five S→I interactions, | ||
| WWW/ | WWW/ | ||
| 53% of 5-word windows: Neither the age group, nor the ethnicity, nor the gender of the account holder had any impact on how SFAs trea… | 53% of 5-word windows: Neither the age group, nor the ethnicity, nor the gender of the account holder had any impact on how SFAs trea… | ||
| - | NDSS/ | ||
| - | 56% of 5-word windows: In the setup in Germany, we can detect the receiver' | ||
| - | NDSS/ | ||
| - | 54% of 5-word windows: In total, our set of discovered WhatsApp servers comprises 410 server instances using 143 different location i… | ||
| - | PETS/ | ||
| - | 57% of 5-word windows: followed by Before Consent (7.57 %) and No Consent (7.34 %).… | ||
| - | PETS/ | ||
| - | 57% of 5-word windows: followed by Before Consent (7.57 %) and No Consent (7.34 %).… | ||
| - | PETS/ | ||
| - | 20% of 5-word windows: No HTTPS was rarest, with 2.85 % of sites.… | ||
| - | PETS/ | ||
| - | 53% of 5-word windows: For a given day 𝑡𝑖, we considered a domain 𝑑 to be problematic if at any point in time within a 7-day windo… | ||
| - | USENIX/ | ||
| - | 56% of 5-word windows: 881 callback numbers were used in more than one campaign, and 2344 callback numbers were used in only one camp… | ||
| CCS/ | CCS/ | ||
| 29% of 5-word windows: We conducted tests using the anti-virus test file EICAR, the exe virus WannaCry, and a PDF virus, respectively… | 29% of 5-word windows: We conducted tests using the anti-virus test file EICAR, the exe virus WannaCry, and a PDF virus, respectively… | ||
| - | IEEE-SP/ | ||
| - | 58% of 5-word windows: Our analysis tracks over 400 phone numbers in 28 countries over the course of 14 months, resulting in a datase… | ||
| - | IEEE-SP/ | ||
| - | 50% of 5-word windows: We found 1.0% of tagged messages across 32 clusters related to advertising.… | ||
| - | IEEE-SP/ | ||
| - | 20% of 5-word windows: VirusTotal returned 417 URLs with at least one detection.… | ||
| - | USENIX/ | ||
| - | 40% of 5-word windows: each reporting service gives a unique ' | ||
| - | USENIX/ | ||
| - | 55% of 5-word windows: Our research finds that the median lifetime of originating sender ID mobile numbers abused to send initial 'hi… | ||
| - | NDSS/ | ||
| - | 53% of 5-word windows: we conducted an empirical study on 29 DNSBL providers by reviewing their websites and public reports, as summa… | ||
| - | NDSS/ | ||
| - | 53% of 5-word windows: We find that 39,201 (76.88%) outgoing mail servers of popular ESPs and websites have been included in at least… | ||
| - | NDSS/ | ||
| - | 50% of 5-word windows: Among these blocklisted domains, the status of 398 (5.67%) domains is serverHold.… | ||
| - | NDSS/ | ||
| - | 56% of 5-word windows: Among the 63 websites that expose their tokens, 56 websites deploy third-party scripts.… | ||
| - | NDSS/ | ||
| - | 50% of 5-word windows: over 80% (47 out of 54) of server and user agent combinations are susceptible to ambiguity-based spoofing atta… | ||
| NDSS/ | NDSS/ | ||
| 58% of 5-word windows: We found that all the VoIP devices are vulnerable to ambiguity-based spoofing attacks to some extent.… | 58% of 5-word windows: We found that all the VoIP devices are vulnerable to ambiguity-based spoofing attacks to some extent.… | ||
| - | NDSS/ | ||
| - | 33% of 5-word windows: real-world deployed services such as Linphone and Linhome are affected by the P-Preferred-Identity preferred a… | ||
| - | NDSS/ | ||
| - | 46% of 5-word windows: all three operators are vulnerable to A4 attack, and one of them is vulnerable to B3 attack.… | ||
| - | NDSS/ | ||
| - | 55% of 5-word windows: We found that 0.8 % of main devices and 13 % of companion devices did not have a one-time prekey at the scan t… | ||
| - | NDSS/ | ||
| - | 57% of 5-word windows: Our analysis across 3.5 B accounts reveals an Android share of 81 % and iOS at 19 %.… | ||
| - | IEEE-SP/ | ||
| - | 50% of 5-word windows: We identify web service spam with 90.78% accuracy (0.87% false positives), with a median feature collection an… | ||
| - | IEEE-SP/ | ||
| - | 56% of 5-word windows: Notably, 77% of initial URL features appearing in email disappear after 15 days.… | ||
| - | IEEE-SP/ | ||
| - | 43% of 5-word windows: We attempted 120 purchases, of which 76 authorized and 56 settled.… | ||
| IMC/ | IMC/ | ||
| 47% of 5-word windows: Table 1: Overview of our smishing dataset collected using posts (n = 220, 585) and image attachments (n = 64, … | 47% of 5-word windows: Table 1: Overview of our smishing dataset collected using posts (n = 220, 585) and image attachments (n = 64, … | ||
| IMC/ | IMC/ | ||
| 18% of 5-word windows: Overall, we collect a dataset with 27.7k smishing messages, 19.3k sender IDs, and 20k URLs.… | 18% of 5-word windows: Overall, we collect a dataset with 27.7k smishing messages, 19.3k sender IDs, and 20k URLs.… | ||
| - | USENIX/ | ||
| - | 54% of 5-word windows: even without any financial incentive, 19.9% of the respondents chose to grant access to their DAB data… | ||
| ============================================================================== | ============================================================================== | ||
| Line 1906: | Line 1812: | ||
| </ | </ | ||
| - | '' | + | '' |
| <file text verify_email_figures-output.txt> | <file text verify_email_figures-output.txt> | ||
| Line 2034: | Line 1940: | ||
| </ | </ | ||
| - | '' | + | '' |
| ===== 13. Guard runs, and what each iteration caught ===== | ===== 13. Guard runs, and what each iteration caught ===== | ||
| Line 2137: | Line 2043: | ||
| '' | '' | ||
| - | A '' | + | A '' |
| - A first version scanned every capitalised token in a 90-character window and reported **eight** false positives on a clean page ('' | - A first version scanned every capitalised token in a 90-character window and reported **eight** false positives on a clean page ('' | ||
| Line 2160: | Line 2066: | ||
| <bibtex bibliography></ | <bibtex bibliography></ | ||
| + | |||
| + | ===== Markup sweep, 2026-09-17 ===== | ||
| + | |||
| + | Mechanical rendering repair only: a fresh live raw/XHTML export of 188 pages was checked with '' | ||
| + | |||
| + | ===== Quote-check refresh, 2026-09-21 ===== | ||
| + | |||
| + | The 2026-09-04 '' | ||
| + | |||
| + | ^ Figure ^ Was ^ Is ^ Why ^ | ||
| + | | detection evidence quotes on the 70 | 400 | 400 | population unchanged | | ||
| + | | verbatim | 220 | 220 | unchanged | | ||
| + | | located after normalisation | 118 | 118 | unchanged | | ||
| + | | rescued from the PDF | — | **49** | these were inside the old 62 | | ||
| + | | below threshold | **62** | **13** (in both renderings) | 62 = 49 + 13 exactly | | ||
| + | | below-threshold share of 400 | 15.5% | **3.3%** | 13 of 400 | | ||
| + | | of those, below 25% coverage | 4 | **2** | 23% and 18% | | ||
| + | |||
| + | **Everything else in the report reproduced byte-for-byte.** A line-by-line diff of the published block against the fresh run differs **only** inside §G. No fold, no per-paper figure, no population count and no external claim on this page or on [[: | ||
| + | |||
| + | **The hand pass and the fallback agree on six of eight, and the two they disagree on are the honest residue.** §8 read eight below-threshold quotes by hand and found all eight present. The fallback reaches six of them on its own. {[englehardt2018_email]} and {[acar2020_noboundaries]} are still below threshold in both renderings, so for those two the August hand read is the only evidence — which is exactly why it was done. | ||
| + | |||
| + | ^ Item ^ Value ^ | ||
| + | | Date | 2026-09-21, unsupervised | | ||
| + | | Command | '' | ||
| + | | Artifact | '' | ||
| + | | Script changes | none | | ||
| + | | Reviewers | one '' | ||
| + | | Pages saved | this page only — [[: | ||
| + | | Open | the 13 have not been read by hand beyond the two named above | | ||
provenance/privacy/email_tracking.1788334860.txt.gz · Last modified: by karel.kubicek.claude