Bytes::Random::Secure CVE-2026-11625

Exploitation requires that the victim application uses Bytes::Random::Secure in one of two specific patterns: an OO generator object instantiated BEFORE the process calls fork(), or the module's functional/procedural interface used across forked processes - in either case the seeded PRNG state is copied into children and never re-seeded. The application must be a multiprocess (forking/preforking) Perl program that generates security-sensitive values from this module, and the attacker must be able to observe or correlate the resulting secrets across processes. Limiting factors: applications that instantiate the generator AFTER forking (per-child), single-process applications, and non-default usage that explicitly re-seeds are not affected; there is no special privilege or authentication needed against the module itself, but realistic impact depends on the secrets being exposed to or usable by the attacker.

Signals are mixed and worth reconciling. The NVD CVSS 3.1 vector (AV:N/AC:L/PR:N/UI:N, C:H/I:N/A:N = 7.5) treats this as a remotely reachable confidentiality issue, which is reasonable insofar as the predictable secrets (tokens, keys) are often exposed or used over the network - but exploitation realistically depends on the victim application using a specific vulnerable coding pattern (object created before fork, or the functional interface), which is a non-default-severity condition the base score does not capture; effective attack complexity is higher than AC:L implies. EPSS is low (0.16%, percentile 6%) and the issue is not in CISA KEV, consistent with no active exploitation and no public exploit code. This is a genuine cryptographic correctness defect with high impact where present, but it is a developer/deployment-pattern-dependent weakness rather than a broadly weaponizable remote vulnerability - prioritize it for Perl multiprocess services (preforking web apps, job workers) that handle this module, but it is not an emergency for environments not running it.

A preforking Perl web application initializes a Bytes::Random::Secure generator at startup, then forks multiple worker processes that each issue session tokens or password-reset values from the shared, identical PRNG state. An attacker who obtains a token from one worker can predict the tokens emitted by sibling workers, hijacking other users' sessions or forging reset links. No public exploit code is identified, but the technique requires only observation and correlation of issued secrets rather than any memory-corruption primitive.