
Dr, apparently. Security Adovcate @semgrep & Hacker. #BugBounty hunter & #infosec YouTuber. APIs & Interlinked OffSec, PhD in AI+Sec @hacknotcrime. she/her
Manchester, UK
Joined February 2018
- Tweets21.8K
- Following1.7K
- Followers1K
- Likes19.2K
Pinned Tweet
Find me on the internet
Mastodon: infosec.exchange/@insiderphd
YouTube: youtube.com/@InsiderPhD
Newsletter: insiderphd.substack.com
Discord: InsiderPhD
Bluesky: bsky.app/profile/insider.phd
LinkedIn: linkedin.com/in/katiepf
Katie Paxton-Fear retweeted
Semgrep Mandoline makes finding triage 10x cheaper while maintaining accuracy. Determinism still works combined with agentic workflows
Java4Shell dropped today and the patching advice is easy: update Java. Working out whether you're actually exposed is not.
Runtime bugs are business as usual now. But this one comes with a twist: it only fires on x86-64 CPUs with AVX-512 VBMI. The same app on the same Java version can have completely different exposure depending on what hardware it runs. Your dependency scanner won't flag it, because it's not a dependency. Your version inventory won't settle it, because the version is just a part of it.
To actually answer "are we affected?" you have to check every JVM, including the ones bundled inside Jira, Jenkins and Keycloak. You need the flags it was really started with, the CPU features that JVM can see, and the node it's running on.. it’s a right mess. Most orgs don't have anywhere close to that kind of view.
IMO This is where agents earn their keep. My colleagues wrote a read-only runbook prompt you can hand to an agent with infra access. It inventories the runtimes, reads the effective flags, checks the CPU, and comes back with a table of affected, mitigated, unknown, and the evidence behind each.
I think this is how incident response goes now. Humans decide yes this matters, and agents do the legwork of finding out what's true across the estate. And the job of a detection engineer is now to figure out scale, agent harnesses, and what they need access to.
Weird times to be living in for sure, check out the full blog:
Katie Paxton-Fear retweeted
Security companies breaking vuln embargos just to be 1st to post aren't helping. They're putting users at risk for SEO. Nobody cares who published first, but we all remember who can't be trusted.
Researchers have loooong memories, we know who you are.
spacerogue.net/?p=1098
For me, at work I was asked by one of my colleagues to relay something someone more senior than me or my colleague.
I was like oh just tell them yourself they’re not going to mind.
And they said they were uncomfortable with it and thought it would be better coming from me, and I had a “oh no I am the adult” moment.
I’m 49. At what point do you feel like an adult and not a child anymore? When does the world make enough sense to give advice to others?
3
12
3,165
Katie Paxton-Fear retweeted
LIMITED TIME: Caido for Hackers Course comes with the Hacking APIs Course FREE.
Just want the API course? Get it for $9.99 (50% off).
Promo ends November 15. Snag one now!
Links in the first reply. 🤓
Claude randomly found a folder on my computer called “research-brain” I use this as a knowledge base but it’s been a while since I’ve updated it with any new content. Claude upon finding this folder has taken it upon itself to resurrect that one particular project and adopt it.
4
22
3,426
Katie Paxton-Fear retweeted
.@monkehack is coming to NahamCon with “Weird Ways to Hack Gemini.”
Ciarán is a bug bounty hunter, and I’m curious to see what made these hacks get labed as "weird". 👀
He takes the stage November 21 in Oakland. Watch the talk live from anywhere at youtube.com/NahamSec.
Katie Paxton-Fear retweeted
where the bounty money went in october
hackerone ≈$1.2M, bugcrowd $551k, intigriti ≈$54k + €26k
pick ur platform before u pick ur target
huntdeck.io/stats
Katie Paxton-Fear retweeted
Find bug -> patch bug doesn't work anymore. We didn't label it Black October for nothing.
Now that AI is writing your code, you have to use AI to find and fix the bugs. The best way to do that is as early in the process as possible.
semgrep.dev/blog/2026/agenti…
Black October is the new normal: your agents write the code, and other agents attack it. That’s why we’re launching the Semgrep Agentic AppSec Platform, with Mandoline, our next-generation program analysis engine purpose-built for agents.
See how it works: semgrep.dev/blog/2026/agenti…



















