Finding and attribution
The Revelio paper identifies a out-of-bounds read in libheif v1.21.2: auxiliary sample-chunk mismatch out-of-bounds read. It lists the author-assessed severity as High and says the finding was manually validated with a reproducible sanitizer-backed proof of vulnerability through a reachable project test harness. This is Table V item 14 of the seven-project real-world hunt.
Product: libheif v1.21.2. AI system: Revelio using Claude Haiku 4.5 and Claude Sonnet 4.6. Team / organization: UC Berkeley Revelio research team and affected project maintainers.
CVE identifiers: CVE-2026-41071.
AI and human contribution (source account): The paper describes Revelio as using Claude Haiku 4.5 for hypothesis generation and Claude Sonnet 4.6 for proof-of-vulnerability confirmation; its experimental-method section says this configuration was used for the seven-project OSS-Fuzz hunt. The authors state all 19 tabled findings were manually validated and sanitizer-confirmed. The paper does not provide a per-item model trajectory, so the exact division of Claude work within this row is not independently stated.
Human involvement is only described where the source account establishes it; unknown involvement does not mean none. Vendor confirmation of a flaw does not confirm every claim about AI contribution.
Discovery context: The paper publicly identifies this distinct finding on 2026-06-20. The authors describe the hunt as searching for previously unknown flaws, but this record does not assert exploitation in the wild or operational zero-day status.
Remediation status: fix documented. The maintainer advisory confirms an out-of-bounds read in SampleAuxInfoReader when saiz sample counts exceed available chunks and lists libheif v1.22.0 as patched.
A source-documented fix does not establish installed patches or current safety. Source checking does not independently reproduce the finding or establish observed exploitation.
Disclosed 19 May 2026. AI attribution dated 20 Jun 2026. Sources checked 14 Nov 2026; these are separate dates, not a live installation check.
Recorded CVSS: Medium · highest recorded score 5.1. Documented technical effects: other.
Severity, consequences and exploitation evidence
Case grouping uses the highest recorded CVSS score, with its version-specific severity band. It is descriptive, not a combined or editorial risk index. All recorded assessments remain below; scores from different versions and sources may disagree. CVSS specifications define version-specific scoring and severity bands.
- CVE-2026-41071: CVSS 4.0 · 5.1 (Medium) ·
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:A/VC:L/VI:N/VA:L/SC:N/SI:N/SA:N. GitHub_M (CNA; CVE Program record) · checked 16 Nov 2026.
CVE-2026-41071: other.
Prerequisites / affected scope: Opening a crafted HEIF file can trigger an out-of-bounds read in the SampleAuxInfoReader constructor. No further consequence is established..
Exploitation evidence: Exploitation in the wild is not established by the cited description.. Impact source · checked 16 Nov 2026.
Documented fix: The maintainer advisory confirms an out-of-bounds read in SampleAuxInfoReader when saiz sample counts exceed available chunks and lists libheif v1.22.0 as patched.