Posts
2528
Following
648
Followers
1466
"I'm interested in all kinds of astronomy."
repeated
Edited 1 year ago
The RockYou2024 compilation (9.95B strings) is so junky that most password crackers are better off just using Hashmob's founds list instead.
Show content

The junk includes:

  • 453M 32-hex hashes
  • 444M digits-only strings of length 8-11 (easily bruteforced)
  • 415M lower-digit or digit-lower strings that are clearly just wordlist words with all possible 4-digit strings appended or prepended
  • 287M of length 6 or less (easily bruteforced)
  • 201M 40-hex hashes
  • 138M bcrypt hashes (plus 15M truncated bcrypts)
  • 71M strings more than 100 characters
  • 51M 96-hex hashes
  • 50M Houzz __SEC__ (modified sha512crypt) hashes
  • 18M encrypted + base64 passwords from the 2013 Adobe leak (credit: Flagg)
  • 12M 32-hex prefixed with '0x'
  • 11M Google auth tokens (ya29 prefix)
  • 7M with at least 20 contiguous hex chars
  • 6.6M 128-hex hashes
  • 160K argon2 hashes

("Easily bruteforced" means that competent attackers are going to run the equivalent hybrid or bruteforce attack anyway much faster on GPU. All these naively-generated strings do is waste attack time ... and inflate the scary size of the compilation 🙄)

If you remove all of this junk (that's useless for directly cracking a human-generated password), all of the RockYou2021 mashup (which was itself similarly problematic), and all founds already available on Hashmob (1.2B) ...

... you're left with only 190M strings that are "net new, maybe useful".

So if you're a pentester or other "normal" password cracker, you can probably just skip RockYou2024. It's only going to be useful if you're a completionist who's trying to crack other mashups (like the long tail of junk in the Pwned Passwords corpus, etc.)

[will update post as I find more non-trivial junk]

3
8
0
repeated

Well, looks like native PDB files finally (unofficially) support compression. A few recent versions of msdia140.dll implement a new MSF format that stores PDB streams in compressed "chunks". It was fairly easy to reverse the implementation, though I have some past experience with the PDB format. Takeaways:

- zstd is used for compression, looks like the open source implementation without any tweaks.
- there doesn't seem to be a way to produce the files in this format at the moment, at least it looks like the code was compiled to only provide deserialization of the new format.
- decompression of chunks is done on-demand. this means that the format is pretty flexible, which allows to optimize for space/speed.(e.g. you can compress the entire stream in a single chunk and get the best compression ratio, but that means the entire stream has to be decompressed at runtime)

I'm gonna write up a converter with some simple compression strategies to see how the new format fares in practice, but that's probably going to take a few days. Hopefully MS comes out with something official soon.

1
3
0
repeated

I finally got up a first draft of docs for ABI Cafe and KDLScript!

https://faultlore.com/abi-cafe/book/

0
1
0
repeated

== Homemade / DIY magnetic tape head, episode 3 ==

As you might have heard, a few days ago we've made a magnetic tape head at home. It is a big deal, because there is a general consensus online that it cannot be done without precise machinery, and if can, will only work on tape pullers working at tape speeds of 38 to 100 cm/s, and perhaps closer to the oldest tape formats with the track width of 1/4" (that should affect volume of the signal).

In this video, the DIY head is playing while using commercial tape puller at 9.5cm/s, and the track is a 1/16" wide (aka single speed domestic standard). This means the frequency response and the signal/noise ratio could be doubled or tripled if we had a faster tape puller.

In this episode:
* Recap of the previous episodes
* Upgrading Mk 1.5 to 1.999
* Erasing the tape at home, and using this head for recording (it works!)
* Ideas for Mk 2

🧵

3
7
0
repeated
Edited 1 year ago

TIL the new YubiKey 5.7 firmware lets Yubico ship keys in a "Restricted NFC" mode, so that folks can't easily talk NFC to them in packaging until they've gotten at least 3 seconds of juice at the destination. Clever. And it can also be toggled by the user!

https://docs.yubico.com/hardware/yubikey/yk-tech-manual/5.7-firmware-specifics.html#restricted-nfc

"Restricted NFC mode prevents wireless device manipulation before a YubiKey NFC with the 5.7 firmware is taken out of its blister pack or other packaging such as a tray. To ensure that these keys cannot be tampered with during shipping, this mode is enabled by default on new NFC keys with the 5.7 firmware.

"When these keys are taken out of their packaging, the only permitted action via the NFC connection is reading the URL configured by Yubico on the NDEF tag set by Yubico. Because both major mobile OSs read NDEF tags and open URLs by default, users immediately learn how to disable Restricted NFC mode. The NDEF tag is set to https://www.yubico.com/getting-started/.

"When tapped against a mobile device, a YubiKey 5.7 NFC will cause the browser to open to the configured URL with the instructions for enabling full NFC operation. The end user is instructed to plug the key into USB power such as a USB charger or computer USB port for 3 seconds. This action is sufficient to disable Restricted NFC mode. The user can re-enable the restriction as often as they desire using the Yubico Authenticator or the YubiKey Manager/ykman."

h/t: Reddit user 'ovirot' - thanks!

1
3
0
repeated
repeated
repeated
repeated
Has anyone checked if macOS is vulnerable to regreSSHion (Qualys's OpenSSH SIGALRM vulnerability, CVE-2024-6387)?

Qualys's writeup notes that glibc is vulnerable because its malloc skips locking on single threaded programs
(https://www.qualys.com/2024/07/01/cve-2024-6387/regresshion.txt),

macOS doesn't have this optimization: malloc always seems to take locks, and will crash in `os_unfair_lock_recursive_abort` if you try to call it during an interrupted call.

So Qualys's initial exploit strategy probably won't work.

macOS's syslog is complex, however; it has several files (https://github.com/apple-oss-distributions/syslog/tree/main/libsystem_asl.tproj/src) that uses malloc, Mach messaging, libdispatch, XPC, ObjC... would there be something non-re-entrant-safe there? If so, would the lock aborts prevent you from exploiting them?
0
4
0
repeated
repeated
repeated
Edited 1 year ago

== Let's make a magnetophone / tape player / magnetic tape head at home! ==

Many people started following me after my DIY magnetic tape and DIY floppy disk experiments. A common request ever since was to make a DIY magnetic head, and, truth to be told, I was curious to experiment with it, too.

The task was daunting, and many people were convinced that it could not be done at all. In fact, I could not find a single mention of a successful experiment in the West, and scarce mentions of it in vintage Russian radio hobbyist magazines. But I know that it could be done; my father says he made some magnetic heads over 40 years ago.

Just two weeks ago Hackaday.com made a post claiming that a (really cool btw) hobbyist made a tape player with a DIY tape head. I was excited at first, and then outraged - it was fake news! The DIY tape head was not (and could not be) used in the tape player on the video, and in fact could only erase tape.

Now, I present you The Real DIY Magnetic/Tape Head (and a DIY microphone)

🧵~

2
8
1
repeated
repeated

did you know that intel shipped a userspace driver that does kernel physical memory grooming (like heap grooming, but for physmem allocations) to get a contiguous memory block https://git.dpdk.org/dpdk/tree/lib/eal/linux/eal_memory.c

like... allocates a bunch of pages, checks if they're physically contiguous, frees the ones that are not, and retries it has enough that are, more or less

2
4
0
repeated
Ok, finally! Here wo go: on via .

Here, we have a basic image running which will be soon shared as a raw image (including instructions how to run it w/o patched and also trying to provide a image.


4
3
0
repeated

🦀 The slides for my workshop at @recon in Montreal this year, "Reversing Rust Binaries: One Step Beyond Strings", are now online!

https://github.com/cxiao/rust-reversing-workshop-recon-2024/tree/main/slides

You can find both the slides and the diagrams I used for the workshop linked there. The slides are meant to be a resource for you to use while reversing, so they have lots of clickable links in them (:

In case you lose the link, you can also find the slides linked from my page on the REcon 2024 schedule: https://cfp.recon.cx/recon2024/talk/QCA37X/

Really great to meet so many cool people, and lots of work to do for Rust RE going forward! I left the conference with a lot of great ideas and directions for new research.

0
7
0
repeated

use-after-free vulnerability due to the interaction between Unix garbage collection (GC) and the io_uring Linux kernel component

https://blogs.oracle.com/linux/post/unix-garbage-collection-and-iouring

Credits Shoily Rahman

0
4
0
repeated

"Saved

MTV News Is Back (Kind Of) Thanks to the Internet Archive

After Paramount Global yanked over 20 years of music journalism, the non-profit Internet Archive created a searchable index of MTV News via its Wayback Machine"

rolling stone.

https://www.rollingstone.com/music/music-news/mtv-news-saved-internet-archive-1235051776/

0
10
0
repeated

Unpatched RCE Vulnerabilities in Gogs: Argument Injection in the Built-In SSH Server https://www.sonarsource.com/blog/securing-developer-tools-unpatched-code-vulnerabilities-in-gogs-1/

0
1
0
repeated

Just released oletools 0.60.2: this is mostly a bugfix release, to address some dependency issues and compatibility with Python 3.12.
More details: https://github.com/decalage2/oletools/releases/tag/v0.60.2
How to upgrade:
pip install -U oletools
or:
pipx install oletools

Another release with new features should come soon!

0
5
0
Show older