Justin Azoff

@JustinAzoff@infosec.exchange
52 Followers
77 Following
131 Posts
Breaking things by looking at them.
Websitehttps://justin.azoff.dev
Githubhttps://github.com/JustinAzoff

This is a really interesting example of why implicit "zero values" for structs in go can be a problem: https://github.com/ubuntu/authd/security/advisories/GHSA-g8qw-mgjx-rwjr

"The temporary user record we returned in the pre-auth NSS request did
not have a GID set, so it defaulted to GID 0 (the ID of the root group)."

I suppose one could argue that it's more of an issue in posix that a uid or gid value of 0 was not defined as nobody/nogroup but root.

New authd users logging in via SSH are members of the root group

### Impact When a user who hasn't logged in to the system before (i.e. doesn't exist in the authd user database) logs in via SSH, the user is considered a member of the root group in the context o...

GitHub

Still not too concerned about the AI robots taking my job.

❯ ollama run gemma3:12b

>>> /clear
Cleared session context
>>> How many letter R's are there in the word strawberry?
There are three "R"s in the word "strawberry".

>>> How many letter R's are there in the word strawberries?
There are four "R"s in the word "strawberries".

All billionaires must submit a list of five things they did for society in the last week or their wealth shall be confiscated.
"Build something small. Own it completely. Scale it thoughtfully. That’s how you win without cheating." https://www.joanwestenberg.com/you-cant-win-in-tech-unless-you-cheat/
You Can't "Win" in Tech Unless You Cheat

PayPal started by paying people to sign up. Their viral growth came from essentially bribing users with $10 to join and $10 for each referral. That’s not innovation — it’s buying market share with venture capital. Uber didn’t disrupt transportation — they ignored taxi regulations their competitors had to

westenberg.

Despite the best efforts of day 22, I managed to get all of #adventofcode 2024 to run in under 100ms (fm13 Ryzen 7840U). This runs and verifies all 49 results including reading and parsing the input files from disk:

❯ ,fast cargo run --release
Finished `release` profile [optimized] target(s) in 0.02s
Running `/dev/shm/aoc2024/release/aoc2024`
Day 1: 0.095ms
Day 2: 0.131ms
Day 3: 0.293ms
Day 4: 0.904ms
Day 5: 0.155ms
Day 6: 4.012ms
Day 7: 4.957ms
Day 8: 0.053ms
Day 9: 4.231ms
Day 10: 0.206ms
Day 11: 2.284ms
Day 12: 1.321ms
Day 13: 0.316ms
Day 14: 7.654ms
Day 15: 0.846ms
Day 16: 7.485ms
Day 17: 0.127ms
Day 18: 1.655ms
Day 19: 0.760ms
Day 20: 3.259ms
Day 21: 0.222ms
Day 22: 34.744ms
Day 23: 14.156ms
Day 24: 6.585ms
Day 25: 0.415ms
----------------
Total: 96.877ms

'cargo flamegraph' was critical as the slowest bits were often not what I thought they were.

Possibly the dumbest thing I've ever written

let val = match value {
"0" => true,
"1" => false,
_ => todo!(),
};

Today's #adventofcode :

How it started:

Day 20: 708429.582ms

How it's going:

Day 20: 4.922ms

I wrote a thing: (Ab)using the NixOS Test framework for clean room PCAP generation: https://justin.azoff.dev/blog/abusing-the-nixos-test-framework-for-clean-room-pcap-generation/
(Ab)using the NixOS Test framework for clean room PCAP generation

Generating clean room PCAP files can be a difficult orchestration problem. The nixosTest framework makes this easy.

If you missed the live version of Justin's webinar yesterday, the stream is online here: https://youtube.com/live/5EatgBUowaw Do you separate out your fiber cables when connecting to your monitoring NIC?
Before you continue to YouTube

I really can't put a finger on why I loathe Intel so much for trying to make me a Lake expert.

8th gen? 12th gen? 13th gen? Bigger number better, sign me right up.

Raptor lake? Coffee Lake? Fuck right off with this bullshit.

Some CPU feature you're looking for was introduced in Ice Lake, and you have a Coffee lake and want to know if you have support for it? Well, no, sorry, there's just no possible way to know something like this without looking up the magic lake decoder ring.