Geoff
sternecker@infosec.exchange
<p>BSidesCLE 2026 | ISC2CLE<br />Cybersecurity | Linux<br /><a href="https://infosec.exchange/tags/InfoSec" class="mention hashtag" rel="tag">#<span>InfoSec</span></a> <a href="https://infosec.exchange/tags/Linux" class="mention hashtag" rel="tag">#<span>Linux</span></a> <a href="https://infosec.exchange/tags/OpenSource" class="mention hashtag" rel="tag">#<span>OpenSource</span></a> <a href="https://infosec.exchange/tags/LGBTQ" class="mention hashtag" rel="tag">#<span>LGBTQ</span></a> <a href="https://infosec.exchange/tags/Vegan" class="mention hashtag" rel="tag">#<span>Vegan</span></a><br />Brain dumps.</p>
Posts
-
Post #4488349
From all the shows I watched as a kid, I thought there would be more quicksand.
-
Post #4436686
"We" are clearly not doing sandboxes and guardrails correctly. Both Anthropic's and OpenAI's models breached test-environment boundaries during UK AI Security Institute evaluations. OpenAI's escaped a sandbox into real Hugging Face infrastructure (something like 17,600 logged actions over 4 days); while Anthropic's used fake identities to reach 3 real organizations from a test environment. These stories made it to NPR. So, mainstream. #AI #Guardrails #Sandbox #InfoSec
-
Post #4221730
Pew pew lasers for space debris.
-
Post #4112091
Phone: Charging rapidly, full in 3 hrs. I don't know that our definitions of rapidly are the same.
-
Post #4111990
Streaming music and a new song is playing, I look up who it is because I like it. It's freaking AI. Damn it. That's not cool, I hoped to learn about a group of actual musicians or even solo artist that I could listen, watch, maybe go see, and no. Hopes dashed. Who is it based off, what prompts, so I can learn more about the sources. sigh filled Sunday #AI #Music #SundaySigh
-
Post #4095193
If my electronics are spying on me.. well, let's just say right now, it's seen some things.
-
Post #4085325
I have an older laptop. I run a lot on it, containers, local AI models. The thing that kills it, aide. :(
-
Post #4062111
RefluXFS (CVE-2026-64600) is a race-condition in the Linux kernel's XFS copy-on-write path. (Kernel updates are available.) On an XFS filesystem with reflink enabled (Default on RHEL and similar, plus Amazon Linux), if you win a race during the copy-on-write remap lets the unprivileged local user overwrite the on-disk contents of any readable file on that volume including /etc/passwd or a SUID-root binary. What makes this wild: the changes persist across reboots, produce no kernel logs, b...
-
Post #4059212
A few things to share, nginx: CVE-2026-42533 dropped yesterday 2026-07-15. It is fixed in nginx 1.31.3, also released yesterday. CVSS 8.1 or 9.2 depending on the nginx version. It&#39;s a heap buffer overflow in the map directive when regex... regardless, it is unauthenticated, needs a custom HTTP request, can crash it or lead to code execution if Address space layout randomization (ASLR) is disabled, or bypassed. Don&#39;t disable memory randomization. Not sure how to bypass ASLR, I&am...
-
Post #4059211
Good News Everyone, BleachBit runs on the SAN! #IT #InfoSec
-
Post #4037525
P vs NP is just Prod || Non-Prod #IT #Infosec
-
Post #4009450
Two previously disclosed CVEs are actively-weaponized kernel root exploits. GhostLock CVE-2026-43499, and Bad Epoll CVE-2026-46242, are both public highly reliable exploits. Ghostlock also appears to enable a container escape and Bad Epoll is also working on Android. Both are fixed in kernel 6.12.96-1. If you have local user, upgrade the kernel & boot it. BRB have to reboot :) #minimalist #Linux #Selfhosting #selfhosted #selfhost #InfoSec #Exploit
-
Post #3965962
440 CVEs dropped to kernel cve-announce in the last 24 hrs?!?!?!? Whoa, calm thy horses! It looks like a backlog dump of resolved issues all sequentially numbered. No RCEs and all resolved. #minimalist #Linux #Selfhosting #selfhosted #selfhost #InfoSec #Exploit
-
Post #3919455
Air quality is bad. A boil notice for water and filter notice because boiling does not remove lead. Ground contamination prevents gardening without raised beds.
-
Post #3910868
A project to test AI guardrails. https://github.com/sternecker/guardrails STRONGER PROSE IS NOT ENFORCEMENT There is a notice in the repo that it is not to be processed. There is a script that will add markdown to the beginning of a PDF. There is an AWESOME original pdf from @pluralistic@mamot.fr and an updated.pdf with the new preface notice.
-
Post #3890265
STRONGER PROSE IS NOT ENFORCEMENT
-
Post #3833013
I listen to the band 7horse quite a bit. The "station" had been playing Cam Cole and I really have been enjoying the music. I had no idea that. Cam Cole is a one man band. Holy crap. Amazing.
-
Post #3820998
Is the TV series silo based on the book wool?
-
Post #3807223
Linus wept, I'm still going through the CVEs.
-
Post #3793797
We lost Sam Neill.
-
Post #3771620
One down a bunch to go One more town another show
-
Post #3766321
some people are harbingers of chaos.
-
Post #3745994
laptop is hung trying to *stop* containers. :/
-
Post #3717644
@radhitya@navi.lain.day wasn't familiar with it, thanks. Looks like it had been idle for a while but has a recent update. cppcheck returned with no critical and a medium for 32-bit builds, which doesn't apply in my case.. Thanks!
- Post #3694664
-
Post #3629311
Attempts to detect/mitigate CVE-2026-46242 "Bad Epoll" Linux kernel, CVSS 7.8, local privilege escalation. https://github.com/sternecker/midil/blob/main/privesc-detection.md
-
Post #3628646
A couple things this morning. CVE-2026-46242 "Bad Epoll" Linux kernel, CVSS 7.8, local privilege escalation was already mentioned CVE published around May. 1: What's new today: it is now fully weaponized and publicly written up. It's a use-after-free race in the kernel's eventpoll subsystem (ep_remove()/ep_remove_file()). An unprivileged local user can win a race condition and get root. 2: Affected: kernel 5.10 through 6.11. 3: The Attack surface is broad because epoll u...
-
Post #3581276
Podman CVE-2026-44517: A breakout can happen during container build using malicious Containerfile and Git Smart HTTP server or GitHub release tar archive. This is weird, essentially it is a path traversal, which has been fixed in 1.43.2. Still pulling malicious code into a container, we can agree, is not ideal path traversal or not. Have not explored the exploitation on this one. #minimalist #Linux #Selfhosting #selfhosted #selfhost #InfoSec #Exploit
-
Post #3581123
PinTheft: CVE-2026-43494, a local root exploit chaining two Linux kernel subsystems. A local unprivileged code execution, needs the RDS/RDS_TCP kernel modules to be loadable, io_uring enabled, a readable SUID-root binary, and just x86_64. Not remotely exploitable, it's a local-root escalation, same class as the DirtyClone family. I previously talked about kernel modules, but assume that an exploit WILL LOAD the modules even if they are not already loaded into the kernel. That being said,...
-
Post #3535580
"Hot rock make steam. Steam make boat go." #QuotesFromWork