GLM-5.3 has helped defend 389 open-source projects, with 4,249 potential vulnerabilities found so far.
OpenVuln is still running. The service remains free, and findings go privately to maintainers.
https://t.co/wqRE3zC2b5
One entropy score can hide the interesting parts.
Map it across a binary to spot abrupt changes and focus your analysis.
https://t.co/aVGiteiBpi
#MalwareAnalysis
A PE file can hide useful data after its last mapped section.
That trailing overlay may contain configs, payloads or noise. Stripping it blindly can destroy valuable data for analysis.
https://t.co/FmwxYnKtHC
#ReverseEngineering#Malware
CVE-2026-88771 - the loaded Citrix footgun went off again, and the screaming noise is back in our ear.
You knew it was coming - enjoy the latest watchTowr Labs blogpost.
https://t.co/cKN1oqLwpK
Interesting take on the next endpoint fight. Everyone is buying into agent security right now and the hard part is that an agent abusing your credentials looks exactly like an agent doing its job.
https://t.co/WyhSuF4E85
#infosec#EDR#AIsecurity
A patch shows more than a fix: its deleted lines can become a search signature for the same bug elsewhere.
That idea became HistRepo: mine BSD fixes, then hunt sibling trees for unfixed code.
https://t.co/iZLZYMvbhr
#BSD#CyberSecurity
Most fuzzers throw strange data at programs.
What happens if you throw strange instructions at the CPU instead?
I revisited an old “trap fuzzing” experiment that found real bugs in OpenBSD and VirtualBox.
https://t.co/Nxq3Oxo4CH
#Fuzzing#CyberSecurity