AndreasChris's recent activity
-
Comment on Vulnerability in OpenSSH on musl Linux distros w/o PAM allows login to password-locked accounts with ANY 257+ character password since 2016 in ~comp
-
Vulnerability in OpenSSH on musl Linux distros w/o PAM allows login to password-locked accounts with ANY 257+ character password since 2016
18 votes -
Comment on YouTube mobile app has deeply hidden the subscription feed in ~tech
AndreasChris LinkHave you tried simply pulling the page back down after it has slid up? Seems to work on android on version 21.43.523. Note that I usually don't use that app to watch YouTube videos, so that's just...Have you tried simply pulling the page back down after it has slid up? Seems to work on android on version 21.43.523. Note that I usually don't use that app to watch YouTube videos, so that's just a snapshot of me testing it a minute ago.
-
Comment on Class of QUIC vulnerabilities across implementations allows QPACK amplification attacks against HTTP/3 servers in ~comp
AndreasChris LinkBecause the last 48 hours were so fun, here's another (less severe) security issue. This time it's the QUIC protocol that's affected. At least we're not dealing with RCE this time, but denial of...Because the last 48 hours were so fun, here's another (less severe) security issue. This time it's the QUIC protocol that's affected. At least we're not dealing with RCE this time, but denial of service is in scope for this one.
A remote, unauthenticated client can force an HTTP/3 server to allocate tens to thousands of times more memory than it sends. One connection is enough to OOM or pin a default-configured server. Seven of the ten implementations tested are vulnerable; several are already fixed.
QPACK compresses HTTP/3 headers to minimize traffic sent over the wire. Across the affected implementations, decoded-header limits are missing, incomplete, or applied too late. A small HEADERS frame can therefore expand into hundreds of thousands of header objects in memory. Preventing that expansion from exhausting resources requires enforcing a cumulative budget as fields are decoded and retained.
The PoC for this one can be found at https://github.com/FoxIO-LLC/blog_poc/tree/main/headeraccordions
It appears the initial report of a related bug was back in 2025, but given that more and more QUIC implementations were discovered to be vulnerable, I guess it took some time for this to be written up and released.
-
Class of QUIC vulnerabilities across implementations allows QPACK amplification attacks against HTTP/3 servers
6 votes -
Comment on Critical GitLab flaw (CVSS 10.0) could allow unauthenticated RCE across twelve years of releases in ~comp
AndreasChris (edited )LinkI thought it couldn't get any worse after yesterday's CA and PHP debacle, but here we go: Yet another RCE vulnerability with a maximum seveeity rating of 10.0. This time in GitLab. At least this...I thought it couldn't get any worse after yesterday's CA and PHP debacle, but here we go: Yet another RCE vulnerability with a maximum seveeity rating of 10.0. This time in GitLab. At least this time the fix is already available, and already deployed on managed installations. Only self-hosted ones may be at risk and have to update immediately.
The issue can be traced back to a flaw in Ruby's builtin regex engine, and affects every GitLab release since Version 7.0. (We're currently at Version 19.5.1 – so that's a looong time.)
But seriously, what is going on? If this continies we're gonna need a vulnerability megathread soon.
-
Critical GitLab flaw (CVSS 10.0) could allow unauthenticated RCE across twelve years of releases
18 votes -
Comment on eMudhra intermediate CA compromise in ~tech
AndreasChris Link ParentAlthough in this case, had you given me this certificate with this prompt and had I found the weak private key, I don't think I would intuitively have suspected this to be a real-world incident...Although in this case, had you given me this certificate with this prompt and had I found the weak private key, I don't think I would intuitively have suspected this to be a real-world incident either (apart from the valid signatures chain to a root cert of course). It just seems to easy for that.
-
Comment on PHP warns it plans to release two emergency patches imminently in ~comp
AndreasChris LinkToday was absolutely wild with regards to IT security. First the CA certificates, then PHP, all withing 24 hours.. :SToday was absolutely wild with regards to IT security. First the CA certificates, then PHP, all withing 24 hours.. :S
-
Comment on eMudhra intermediate CA compromise in ~tech
AndreasChris Link ParentThis whole story kinda feels like someone read about the PRNG used by the NSA to generate the ECDSA seeds and thought "Hold my xkcd, I can top that."Anyway, what the actual fuck, how is one of the private keys "4"? (meme, and somehow a relevant xkcd)
This whole story kinda feels like someone read about the PRNG used by the NSA to generate the ECDSA seeds and thought "Hold my xkcd, I can top that."
-
Comment on eMudhra intermediate CA compromise in ~tech
AndreasChris Link ParentThat is absolutely wild. In recent months I couldn't help but wonder wether the whole IT security job market will be negatively impacted by new LLMs discovering high volumes of increasingly...That is absolutely wild. In recent months I couldn't help but wonder wether the whole IT security job market will be negatively impacted by new LLMs discovering high volumes of increasingly complex bugs in the long term. Then I see stuff like this. If that isn't IT security 101, I don't know what is...
-
Comment on Sweden's national broadcaster SVT suspects some of the 4,077 songs submitted for Melodifestivalen 2027 were generated using artificial intelligence in ~music
AndreasChris Link ParentIn the context of the ESC the song needs to be performed by a human live on stage in the end anyway. So I suppose those AI submissions are songs originally generated by AI and re-recorded by a...In the context of the ESC the song needs to be performed by a human live on stage in the end anyway. So I suppose those AI submissions are songs originally generated by AI and re-recorded by a human performer, are they? I can see how you wouldn't be able to tell these apart from a song originally conceived by a human from the recording alone.
-
Comment on Origin of ECDSA curve seed provided by NSA to NIST in 2000 without explanation turns out to be simple PRNG output with UNIX timestamp as seed in ~comp
AndreasChris LinkIn case you don't know what this is about: Simply put ECDSA is a signature algorithm standardized by NIST in 2000 and used in all sorts of places to secure digital communication in day-to-day...In case you don't know what this is about:
Simply put ECDSA is a signature algorithm standardized by NIST in 2000 and used in all sorts of places to secure digital communication in day-to-day applications. Hence the ability to break it, would have huge implications for IT security.
The algorithm uses a magic seed value that has been provided by the NSA (who generated the curve parameters) without explanation. Over the years this lead to a number of conspiracy theories, that the seed may have been chosen to contain a secret backdoor.
One day an anecdote surfaced, that the seed may have simply been a SHA-1 hash of some english sentence chosen by a NSA employee. Huge efforts have been made over the years to find out how the seed may have been generated, and there was a five figure cash prize to be paid out to the first person to find it.
Then yesterday, anthropic posted this blogpost, announcing that some research model of theirs found the actual origin of the seed. Turns out it was just the output of several iterations of a standard PRNG from some 1990s C library seeded with a timestamp. I.e. as basic as it gets when it comes to generating pseudo random numbers.
This mostly debunks the conspiracy that the seed is part of some grand conspiracy to backdoor ECDSA, so good news I guess. But for the most part I can't help but find it very funny how a large online community spent years and loads of resources hashing random english sentences in the hopes of finding the seed's origin, only for an LLM to reveal it's been the output of a non-cryptographically-secure pseudo random number generator from an old C library all along. :D
Or in other words (joke): https://www.youtube.com/watch?v=eOQfNT6Jdkg
-
Origin of ECDSA curve seed provided by NSA to NIST in 2000 without explanation turns out to be simple PRNG output with UNIX timestamp as seed
34 votes -
Comment on Jet Lag Season 19: Japanorama | Trailer in ~hobbies
AndreasChris Link ParentDunno, the pacing of... ...the final ten minutes of the episode was quite well edited in my opinion, so I don't really mind the 'early' conclusion. I found it much more frustrating how some veeery...Dunno, the pacing of...
...the final ten minutes of the episode was quite well edited in my opinion, so I don't really mind the 'early' conclusion. I found it much more frustrating how some veeery lucky last-minute draws kinda decided this season on pure chance. After this much back-and-forth I really would've preferred a tie in this one.
-
Comment on Jet Lag Season 19: Japanorama | Trailer in ~hobbies
AndreasChris Link ParentI've only watched the beginning of episode 8 so far, but the first few minutes... ...went as I expected/feared. It would've really surprised me if they did a random draw, actually lost their...I've only watched the beginning of episode 8 so far, but the first few minutes...
...went as I expected/feared. It would've really surprised me if they did a random draw, actually lost their Shinkansen card, and still made the penultimate episode end on that cliffhanger. -
Comment on Jet Lag Season 19: Japanorama | Trailer in ~hobbies
AndreasChris LinkEpisode 8 (finale) is now on Nebula -
Comment on Jet Lag Season 19: Japanorama | Trailer in ~hobbies
AndreasChris Link ParentEpisode 7 is now on YouTube -
Comment on Jet Lag Season 19: Japanorama | Trailer in ~hobbies
AndreasChris Link ParentYeah, they already 'apologized' for... ...the cliffhanger in the layover podcast xD: But yeah, that ending was absolutely wild. I kinda want to believe "Team Fight" makes a miracle-comeback and...Yeah, they already 'apologized' for...
...the cliffhanger in the layover podcast xD:
Ben: And look, we formally apologize for this cliffhanger. It was to good to not do. We had – we had to do it.
Sam: Right, we get ooone –
Ben: We get one mean cliffhanger, and its before the finale, and its to good.But yeah, that ending was absolutely wild. I kinda want to believe "Team Fight" makes a miracle-comeback and manages to win, but given how much content is left my feeling tells me that Sen&Bam won't lose their Shinkansen card prematurely.
-
Comment on Jet Lag Season 19: Japanorama | Trailer in ~hobbies
AndreasChris Link ParentEpisode 6 is now on YouTube
Basically, on linux distros with the musl C library in use, the library function crypt() does not accept 257+ character passwords. On failure, the return value is * instead of an error.
In 2016, OpenSSH started to hash passwords for login attempts, even if an account has no real password hash. To keep the timing of responses consistent (in order to prevent sidechannel attacks), other accounts' salts are used in these cases.
For Passwords longer than 256 characters, this means we get a return value of *, which matches the value * stored in /etc/shadow for root, key-auth-only users, and some other cases. Hence the comparison performed by SSHD succeeds and the login attempt is let through.
Note that only sshd with password auth enabled and w/o PAM is affected. It is patched in OpenSSH 10.6p2. So make sure to update, or at least turn off password based authentication as a workaround.