Attackers are using a password authentication bypass in OpenSSH on musl-based Linux to log into accounts whose password field is set to an asterisk, then hunting for keys, hashes and cloud metadata.
A jitterentropy health test run in FIPS mode during early boot intermittently fails on certain AMD EPYC processors, and the kernel panics instead of starting.
OpenAI says the model beats GPT-5.6 Cyber at exploitation while staying in scope more reliably, and reports 304 critical issues to maintainers with proposed fixes.
Previdian says attackers went after the musl-without-PAM password bypass quickly, and one honeypot gave up root. The bug only bites a narrow set of builds.
On Alpine and other musl libc builds with PAM disabled, some accounts accept any password between 257 and 1,024 characters. Fixes are in OpenSSH 10.6p2.
XLab found a fake updater script that swaps the legitimate cwpsrv binary for a malicious one, then waits for an encrypted HTTP header before running anything.
A password of 257 to 1,024 bytes gets past SSH authentication on some Alpine servers running the standard openssh-server package, including accounts with no password set.
An unreleased OpenAI cyber model ran an autonomous pipeline against the kernel's master branch and produced working local privilege escalation exploits, at a cost of $1.2 million in API credits.
A researcher says the bug has been in OpenSSH since 2016 and only bites on musl-based systems with password auth on and PAM off. It's fixed in OpenSSH 10.6p2.