radare2 CVE-2026-81882

Exploitation requires local use of radare2 (< 6.2.0) by an analyst who (1) loads attacker-controlled binary property-list data and (2) explicitly issues the pFB or pFBj print command against it - the flaw is not reachable in normal analysis without running these specific commands. Strong limiting factors: no network vector (AV:L), mandatory user interaction (UI:R - the operator must run the command on hostile input), no authentication bypass or privilege gain, and impact bounded to leaking uninitialized/adjacent heap bytes into JSON output or crashing the radare2 process. There is no code execution or write primitive.

This is a genuinely low-priority issue, and the signals are internally consistent rather than conflicting. The vendor CVSS 3.1 vector (AV:L/AC:L/PR:N/UI:R/S:U/C:L/I:N/A:N, base 3.3) correctly reflects that exploitation is local and requires the victim analyst to actively load untrusted binary property-list data AND explicitly run the pFB/pFBj print commands (UI:R). There is no network vector, no privilege escalation, and no memory corruption/write primitive - CWE-125 here is an out-of-bounds heap READ whose worst outcomes are disclosure of uninitialized/adjacent heap bytes into JSON output (C:L) and a possible crash. No EPSS score, no CISA KEV listing, and no public exploit are present in the input; the two blocks in the fix commit are regression tests, not weaponized POC. The main tension worth flagging: the description explicitly says 'possible process termination,' which supports A:L rather than the vendor's A:N, so availability impact is slightly understated. Overall this is a low-real-risk defect in a specialist analyst tool, not an urgent remediation priority; patch during normal maintenance by upgrading to radare2 6.2.0.

A reverse engineer opens an attacker-supplied binary property-list file in radare2 and runs the pFB or pFBj command to pretty-print/JSON-encode it. Because the Unicode parser underallocates the UTF-8 destination buffer and leaves it without NUL termination, the JSON encoder reads past the allocation, leaking uninitialized or adjacent heap contents into the printed JSON (or crashing the process). No public exploit is identified at time of analysis, and exploitation depends entirely on the analyst voluntarily running the command on hostile data.