Pick what you run. Lists the CVEs Apple fixed after that version — in later updates of the same version line, and fixes that only shipped in newer major versions. Based on Apple's advisories only; a CVE fixed only in a newer major version may or may not affect older ones.
CVEs with a known “reported to Apple” date — from the Project Zero issue tracker, ZDI advisories’ disclosure timelines, public WebKit bugs and write-up timelines (only a fraction of all CVEs). Days are counted to Apple’s first published fix. All filters apply.
About this data
Sources, and how far we trust them
| Source | Used for | Trust |
|---|---|---|
| Apple security releases and Apple's archive pages | Everything about a CVE: release, date, component, “Available for”, impact, description, credit, “may have been exploited”, “Entry added”. Pages Apple deleted (2005–2009) are restored from Wayback Machine snapshots; 2002–2004 come from Apple's two archive pages. | Primary. Taken as published; two errors in Apple's own index pages are corrected. |
| NVD | CVSS base score only (3.1, then 3.0, 4.0, 2.0). Nothing else from NVD is used. | As published by NVD. Many recent CVEs have no score yet. |
| CVE.org records | Who assigned a CVE (CNA), when the record was published and the optional “made public” date: third-party lag. | Good for the CNA. Record dates can be days to years after the bug became public; see the warning on the Third-party lag page. |
| Project Zero tracker, ZDI advisories, WebKit Bugzilla, write-ups | “Reported to Apple” dates (Fix timelines). Order of preference: Project Zero's “Reported” field, ZDI's disclosure timeline, the date a public WebKit bug was filed, timeline lines in write-ups. | Structured fields are reliable. Every date must be before Apple's first fix and at most 3 years before it. ZDI dates are dropped when ZDI published long before Apple's fix (0-days, or a CVE id added later). |
| Public research: PoC-in-GitHub, trickest/cve, Exploit-DB, GitHub search, ~55 blogs and conferences, Speaker Deck, YouTube | Links to PoCs, write-ups, slides and talk recordings. Only links and titles are stored. | Heuristic. A link only gets a tag when it is clearly about the CVE. GitHub repos are checked for code: README-only repos are dropped, binary-only repos are flagged “suspicious”. PoCs are unvetted third-party code. |
| Apple open-source releases | Source-diff links for macOS fixes (for example xnu 15.3 → 15.4): the project tags pinned in each macOS release. | Exact. The diff covers every change between the two releases, not only the fix. |
| Curated by hand | In-the-wild campaigns (each CVE checked against the Citizen Lab / Amnesty / Kaspersky / Google page that names it); minimum Macs per macOS version (Apple's compatibility pages); jailbreaks and their CVEs (The Apple Wiki); reporter name merges and affiliation aliases. | Reviewed, but small and maintained by hand: may lag behind. |
How the data is processed
- Parsing. Every advisory page is cached and parsed into releases, entries and CVEs. Platforms and exact versions come from the release title (“iOS 17.5 and iPadOS 17.5”); combined macOS releases also name older versions in their “Available for” lines.
- Reporters. Credit strings are split into people, affiliations (“of Google”) and disclosure programs (“working with Trend Micro ZDI”). The same person written differently is merged by name, alias and handle; ambiguous cases are kept apart. The original credit text is always shown.
- Impact classes (code execution, privilege escalation, …) are keyword rules on Apple's impact text.
- Exploited means Apple's own wording (“Apple is aware of a report that this issue may have been exploited”). The In the wild page also counts CVEs named in a curated campaign.
What the numbers mean
- First fixed: the date of the earliest Apple release listing the CVE. Patch lag: days from that to the fix on each other platform.
- Added later: Apple added the entry to the advisory after the release (“Entry added …”).
- Report → fix: days from the report date (above) to Apple's first fix.
- Patch coverage: an older version counts as supported for a fix if Apple shipped a regular security update for it (one fixing 5+ CVEs) on that day or within 150 days after, or if it got that fix anyway. One-off updates for abandoned versions (iOS 12.5.8, 16.7.16, Big Sur 11.7.11) don't count. For macOS, a Safari update that installs on an older version counts as a fix there. watchOS, tvOS and visionOS aren't included: Apple doesn't maintain their older versions. “Not fixed” means no fix was published: Apple doesn't say whether that version is affected.
- Device support: device minimums come from “Available for” (Mac: Apple's compatibility pages). “Still updated” = a regular security update in the last 120 days; “one-off fixes only” = Apple still ships the occasional small fix; “ended” = nothing for a year.
- Third-party lag: days from when a third-party bug became public to Apple's fix. It's an approximation; see that page.
- Collisions: CVEs credited to people from different organisations, usually independent discoveries.
How it is checked
- A cross-check opens a sample of Apple's live pages in a browser and compares them with this data (CVEs, components, credits, platforms): currently 99.7% of about 20,000 checks agree.
- Automated tests check the data's integrity, known answers from past fixes and that every export agrees, plus about 100 checks of the site itself.
- Updates are run by hand, usually after each Apple release. The date at the top shows the last update.
Downloads and reuse
CSV (one row per CVE), JSON per year (api/), the SQLite database, Atom feeds and one page per year (cve/2024.html#CVE-…) are linked in the footer.
The site only links to third-party research; nothing is copied or hosted. This product uses the NVD API but is not endorsed or certified by the NVD.
The visual design is inspired by holadelej.hu.