OpenBSD PPP PAP Flaw Allowed Authentication Bypass Across All Releases
Vulnerability Disclosure · 7 sources · Updated 5mo ago · First seen Jun 16, 2026

OpenBSD fixed a high-severity authentication bypass in its PPP stack after researchers disclosed that the bug had existed since the original sppp code import from FreeBSD in 1999. The flaw affects sppp_pap_input() and all OpenBSD releases through 7.6 when the system is configured as a PAP authenticator on the PPPoE data path. By trusting attacker-controlled name_len and passwd_len values in bcmp() comparisons, the code could accept zero-length credentials as valid because a zero-length bcmp() returns success, allowing remote attackers to authenticate without legitimate credentials.
The disclosure also described a secondary kernel heap over-read caused when a supplied username length exceeds the allocated size of the stored credential, a condition introduced after credential fields became dynamically allocated in 2009. Researchers said the bug could let an attacker intercept or read PPPoE traffic and could also allow a rogue PPPoE server to impersonate a legitimate server during mutual authentication. OpenBSD committed a fix to -current on 2026-06-14, and a proof of concept reportedly showed successful PAP authentication with empty credentials followed by establishment of a full network-layer link.


