DomainTool's malware over DNS inspired me to finish this proof-of-concept game utilizing DNS TXT records dynamically generating Mad Libs in JavaScript. Think Cloudflare pastebin-over-DNS.
https://ttl.ninja/madlibs/madlibs.html
Web | https://badsamurai.dev |
GitHub | https://github.com/BadSamuraiDev |
Verification | https://www.badsamurai.dev/mastodon-verification |
Gravatar | https://gravatar.com/badsamuraidev |
Pronouns | he/him |
Dogs | An old clever "mini" golden and blind black lab-ish pup |
DomainTool's malware over DNS inspired me to finish this proof-of-concept game utilizing DNS TXT records dynamically generating Mad Libs in JavaScript. Think Cloudflare pastebin-over-DNS.
https://ttl.ninja/madlibs/madlibs.html
Today I was pondering my alt resume of some generally useless tech skills (typically admin + dev)
I'd like to say we learned these only to later build something better. But we sure liked a lot of terrible products at the time.
I am not the best at domain intel, or OSINT, or anything really.
But what I want spammers and scammers and botters to understand is if they target Mastodon in general, or infosec folks here specifically, I will grumpily spend a large chunk of my free time finding out who you are and/or interfering with your fuckery.
I do not do work/life balance, and I have great datasets at hand.
Done digging. Pretty confident in the identity, ambivalent about wreaking havoc on this dummy but happy to hand off details to @hacks4pancakes@infosec.exchange @luckytran@med-mastodon.com or @nixCraft@mastodon.social I sent a polite email off to the young gentleman it appears to be. Hopefully he realizes his mistake.
Knostic sponsored a "Prompt Pit" this morning - a gaggle of infosec folks got together to share quick GenAI-related findings, tips, or use cases.
Dragos Ruiu introduced a term I haven't heard before that could've come straight from @bruces:
"context rot."
When working with an LLM you get to a point where the past text becomes too large for the LLM to consider, moving outside the model's "context window" and degrading generation with respect to the whole conversation.
https://docs.anthropic.com/en/docs/build-with-claude/context-windows
Since DNS is on π₯ today I should note if you're a Splunk shop, the DNS data model in Enterprise Security does not include the field for TXT record values, you need to add that manually.
Then you can do high-fidelity detections such as length and base64 with conversions looking for code.