Private researcher reported a security vulnerability to Chromium on August 4. The fix went into the public source code but Chrome users never got it due to patching timelines. Two Chinese groups, apparently read that public fix, built an exploit chain from it and started phishing NGOs on September 1 using identical code. Technically an N-day but for anyone actually running Chrome, a zero-day. Here the public patch knowledge worked as an instruction manual. Open source means everyone can read the fix. With AI, some people just operationalise the knowledge faster. https://t.co/1pY3F4qCl9
🤓 At @UnpromptedAU I presented some of the experimentation I did with @typesafeai Jev.
I used Jev to evaluate AI agent drift and misalignment to detect if my agent is going rogue or being compromised.
Jev is a model used for data classification and decision-making.
It uses 3 different primitives:
Noul: Is the agent off-task?
Score: How severe is the drift?
Choice: What behavior is it?
You can define your own criteria depending on the data you are analyzing.
It is cheaper and faster than an LLM, and the data you classify can be anything, which makes it great for threat intel and cybersecurity!
In my example below, Jev is evaluating if my agent is being compromised through a prompt injection loaded from external data and passed into the context of my agent.
I also created a utility to analyze my codex and claude code agent sessions and discover if something went wrong or was unusual during the session.
I can see a lot of opportunities for threat detection and data classification. I still need to experiment more but it opens new possibilities.
More to come 🧐
Why does SCCM app execution spawn from WmiPrvSE.exe instead of CcmExec.exe?
@Praga_Prag traced the full chain with ProcMon + Ghidra, from WMI tasking to CreateProcessW, and built a detection strategy around it.
Check it out! https://t.co/OO9YEOhePF
Token Theft in Microsoft Entra ID, by @insinuator
Part 1: Threat Landscape and Attack Techniques https://t.co/jIgt0hXkU5
Part 2: Continuous Access Evaluation https://t.co/F6SepUgByd
⚡ Fine-tune and quantize AI models directly on NVIDIA Jetson.
Our new Jetson AI Lab tutorial shows how to use Unsloth and memory-efficient QLoRA to customize models, export quantized GGUF files, and run them locally with llama.cpp.
Follow hands-on examples for:
🔹 NVIDIA Nemotron 3.5 Lightning on Jetson AGX Thor
🔹 Qwen3.5-4B on Jetson Orin Nano
Start optimizing: https://t.co/hDUCiNnQcW
There are numerous ways to extract cloud credentials (like tokens) once an attacker gains access to a workstation. Protect your cloud admin credentials as Tier 0.
https://t.co/yjD8tKv2eA
One curl command on an exposed Azure VM poses a risk of a full subscription compromise via IMDS.
Managed Identity tokens are the ultimate backdoor.
https://t.co/65VfBQWqYB
Blog: "Bad advice and myths around malware prevention"
If you ever heard or said "visiting websites can't infect you", "PDFs aren't malicious" or "exploits are rare and always targeted" this article might be for you.
https://t.co/LYpRaaEsSG
#GDATATechBlog#GDATA
If you want a quick primer on passkeys, how they are phishing resistant, etc., check out the hacker summer webinar I did with @AlteredSecurity
There are new considerations with passkeys, but they're orders of magnitude better than passwords and MFA
https://t.co/wqaUiB6Q4H
Did you know Entra has recommendations for configuration settings and can even send email alerts on expiring app/service principal credentials?
Ideally, I recommend using Maester which has far broader coverage, but this is a nice built-in feature :)
https://t.co/zORKy7WZpv
My analysis from hosting Hugging Face at the CISO huddle for the Cloud Security Alliance, operational to program building.
Remember: Agents… find a way.
A thread
Sigma rules outdetect GuardDuty in AWS post-compromise testing, catching 11 of 32 simulated attacks vs. GuardDuty's 3, pointing toward a Chainsaw-style first-response tool for CloudTrail logs.
Key findings:
- Invictus IR and @BertJanCyber ran 32 attack simulations via Stratus Red Team against a live AWS environment, logging everything to CloudTrail streamed to S3 and queried through Athena. GuardDuty fired on only 3 of those 32 well-known techniques, a significant gap for a paid managed detection service.
- Sigma covered 11 of 32 simulations, nearly 4x GuardDuty's coverage, but the Sigma-to-Athena conversion via https://t.co/LH21YluWKN breaks field mappings: eventSource and eventName are mistranslated to nonexistent fields like https://t.co/jGoC0Y8JHN. The team built a fixup script to patch those queries automatically before execution.
- The gap to 100% Sigma coverage is partly structural: current Sigma rules lack aggregation support (e.g. ConsoleLogin greater than 10), rules are fragmented across repos making quality assessment hard, and cloud-specific rules are vendor-siloed rather than threat-centric.
- The vision is a Chainsaw equivalent for cloud: a tool that runs translated Sigma rules against CloudTrail logs in Athena, usable against live environments or logs imported into an isolated AWS account for safe analysis.
Takeaway: The Sigma rules, Athena and Splunk queries, field-fixup script, and the full 32-attack dataset are on their GitHub.
#DFIR_Radar
RID-500 is AD's best break-glass account... and you should disable it!🙃
@JsQForKnowledge shared counterintuitive wisdom at #TROOPERS26
- All-powerful and cannot get locked out.
- Disabled = more secure.
- Disabled bypassed when booting in Safe Mode With Networking.
Sources in🧵
A 1.5B security model that runs locally serious Blue/Red team reasoning on 1.2 GB to 4GB RAM with CPU.
Fine-tuned from DeepSeek-R1-Distill-Qwen-1.5B using,
covering :
- Prompt injection & tool poisoning detection
- Ransomware incident response playbooks
- MITRE ATT&CK mapping + detection rule generation
- CVE/CWE reasoning
- Financial fraud pattern analysis
chain-of-thought +135% reasoning score, 100% <think> block activation, and practical coverage of threats all in a 1.5B model that fits on weak hardware.
Built for pentesters, detection engineers, and security researchers who need offline capability.