An Australian law firm detected anomalous behavior involving two employees accessing numerous sensitive files, which triggered alerts via User and Entity Behavior Analytics (UEBA). Investigation revealed signs of lateral movement and privilege escalation attempts, indicative of a potential ransomware attack. The firm successfully contained the threat before any encryption or data exfiltration occurred, demonstrating the effectiveness of behavioral analytics and least privilege enforcement. This case highlights the importance of monitoring deviations from normal access patterns to detect early stages of ransomware and insider threats. No actual breach or payload delivery took place. The threat underscores risks related to privilege escalation and lateral movement within hybrid environments. It serves as a practical example of proactive defense mechanisms preventing ransomware disasters. European organizations with similar environments and sensitive data face comparable risks. Early detection and strict access controls are critical to mitigating such threats.
I wrote a hands on guide that shows how leaked webhooks surface as an attack vector; how to find them in the wild; how to craft safe non destructive PoCs; how to harden receivers. Includes curl examples for Slack and Discord; Node.js and Go HMAC verification samples; a disclosure template.
Why this matters
* webhooks are often treated as bearer secrets; leaks are common
* small mistakes in verification or ordering can become business logic bugs
* many real world impacts are serviceable without flashy RCE
What you get in the post
* threat model and scope guidance
* detection rules and SIEM ideas
Read it here: [https://blog.himanshuanand.com/posts/2025-09-17-how-to-hack-webhooks/](https://blog.himanshuanand.com/posts/2025-09-17-how-to-hack-webhooks/)
Notes: do not test endpoints you do not own.follow program scope and responsible disclosure rules.
Happy hunting
With the recent npm/Node.js supply chain incident (phished maintainer, 18 packages briefly shipping crypto-stealing code), I wanted to share a small project:
**Typo squat Detective,** a 2-3 minute browser game to practice spotting look-alike domains.
It covers:
• Numbers ↔ letters (1 ↔ l, 0 ↔ o)
• Unicode homoglyphs (Cyrillic/Greek lookalikes)
• Punycode (`xn--`) tricks
Play it here: [https://typo.himanshuanand.com/](https://typo.himanshuanand.com/)
Curious to hear which tricks fooled you and if you would like more levels/brands.
*Today, we have intuitively adapted to Quick Response (QR) codes in the tapestry of our day-to-day life.*
Whether we are scanning the menu at a restaurant or checking in to our flight, they have seamlessly integrated into the consumption of products and information.
Yet, with its widespread adoption, it has also acquired an accompanying security challenge. Bad actors have capitalized on QR codes to perpetrate phishing attacks, deliver malware, and exfiltrate data without the user’s knowledge.
Oftentimes, this is caught after the attack occurs. This is why you should [use a safe online QR code scanner](https://www.qrstuff.com/scan) that authenticates the link before you click on it is so critical.
Just published a breakdown of RapperBot. Quick hits:
Uses DNS TXT records to hide rotating C2s.
Multi-arch payloads (MIPS, ARM, x86), stripped/encrypted, self-deleting.
Custom base56 + RC4-ish routine just to extract C2 IPs (decryptor included).
Infra shifts fast: scanners moving countries, repos/FTP/NFS hosting binaries.
Timeline lines up neatly with DOJ’s Operation PowerOFF takedown.
Full post: [https://www.bitsight.com/blog/rapperbot-infection-ddos-split-second](https://www.bitsight.com/blog/rapperbot-infection-ddos-split-second)
Curious if anyone’s still seeing RapperBot traffic after the takedown, or if it’s really gone quiet.
A critical flaw in the [Mobile Security Framework (MobSF)](https://cybersecuritynews.com/mobsf-framework-zero-day-vulnerability/) has been discovered, allowing authenticated attackers to upload and execute malicious files by exploiting improper path validation.
An interactive application that visualizes and demonstrates Google’s CaMeL (Capabilities for Machine Learning) security approach for defending against prompt injections in LLM agents.
Link to original paper: [https://arxiv.org/pdf/2503.18813](https://arxiv.org/pdf/2503.18813)
All credit to the original researchers
title={Defeating Prompt Injections by Design},
author={Edoardo Debenedetti and Ilia Shumailov and Tianqi Fan and Jamie Hayes and Nicholas Carlini and Daniel Fabian and Christoph Kern and Chongyang Shi and Andreas Terzis and Florian Tramèr},
year={2025},
eprint={2503.18813},
archivePrefix={arXiv},
primaryClass={cs.CR},
url={https://arxiv.org/abs/2503.18813},
}
In a recent red team engagement, the client's attack surface was so well-defended that after months of effort, the only system we managed to compromise was a lone server, which was apparently isolated from the rest of the network. Or so we thought.
One developer had been using that server for remote development with Cursor. This setup is becoming increasingly popular: developers run AI agents remotely to protect their local machines.
But when we dug deeper into how Cursor works, we discovered something unsettling. By pivoting through the remote server, we could actually compromise the developer's local machine.
This wasn't a Cursor-specific flaw. The root cause lies in the Remote-SSH extension that Cursor inherits directly from VS Code. Which means the attack path we uncovered could extend across the entire VS Code remote development ecosystem, putting any developer who connects to an untrusted server at risk.
For the details, check out our [blog post](https://blog.calif.io/p/vibe-hacking-abusing-developer-trust). Comments are welcome! If you enjoy this kind of work, [we're hiring](https://calif.io/jobs)!
Can you review my new security testing tool [https://github.com/MohamedKarrab/odoomap](https://github.com/MohamedKarrab/odoomap)
Features:
• Detect Odoo version & exposed metadata
• Enumerate databases and accessible models
• Authenticate & verify CRUD permissions per model
• Extract data from chosen models (e.g.res.users, res.partner)
• Brute-force login credentials (default, custom user/pass, wordlists)
• Brute-force internal model names when listing fails
Lets imagine a scenario where you're coerced whether through threats, torture, or even legal pressure to reveal the password to your secure vault.
In countries like the US, UK, and Australia, refusing to provide passwords to law enforcement can result months in prison in certain cases.
I invented a solution called [Veilith](https://www.veilith.com/) ( [veilith.com](http://veilith.com/) ) addresses this critical vulnerability with perfect deniable encryption. It supports multiple passwords, each unlocking distinct blocks of encrypted data that are indistinguishable from random noise even to experts. And have a lot of different features to protect your intellectual properties.
In high-stakes situations, simply provide a decoy password and plausibly deny the existence of anything more.
Dive deeper by reading the whitepaper, exploring the open-source code, or asking me any questions you may have.
Just published a new write-up where I walk through how a small HTTP method misconfiguration led to admin credentials being exposed.
It's a simple but impactful example of why misconfigurations matter.
📖 Read it here: [https://is4curity.medium.com/admin-emails-passwords-exposed-via-http-method-change-da23186f37d3](https://is4curity.medium.com/admin-emails-passwords-exposed-via-http-method-change-da23186f37d3)
Let me know what you think — and feel free to share similar cases!
\#bugbounty #infosec #pentest #writeup #websecurity
\#️⃣ How we Rooted Copilot #️⃣
After a long week of SharePointing, the [Eye Security](https://www.linkedin.com/company/eyesecurity/) Research Team thought it was time for a small light-hearted distraction for you to enjoy this Friday afternoon.
So we rooted Copilot.
It might have tried to persuade us from doing so, but we gave it enough ice cream to keep it satisfied and then fed it our exploit.
Read the full story on our research blog - [https://research.eye.security/how-we-rooted-copilot/](https://research.eye.security/how-we-rooted-copilot/)
Mateusz Lewczak explains how the AFD.sys driver works under the hood on Windows 11. In Part 1 [1], he demonstrates how to use WinDbg and the NtCreateFile call to manually craft a raw TCP socket, bypassing the Winsock layer entirely.
Part 2 of the series [2] dives into the bind and connect operations implemented via AFD.sys IOCTLs. Mateusz shows how to intercept and analyze IRP packets, then reconstruct the buffer needed to perform the three‑way TCP handshake by hand in kernel mode.
[1] https://leftarcode.com/posts/afd-reverse-engineering-part1/
[2] https://leftarcode.com/posts/afd-reverse-engineering-part2/
I recently investigated a Red Bull-themed phishing campaign that bypassed all email protections and landed in user inboxes.
The attacker used trusted infrastructure via post.xero.com and Mailgun, a classic living off trusted sites tactic. SPF, DKIM and DMARC all passed. TLS certs were valid.
This campaign bypassed enterprise grade filters cleanly... By using advanced phishing email analysis including header analysis, JARM fingerprinting, infra mapping - we rolled out KQL detections to customers.
Key Takeway: No matter how good your phishing protections are, determined attackers will find ways around them. That's where a human-led analysis makes the difference.
Full write-up (with detailed analysis, KQL detections & IOCs)
https://evalian.co.uk/inside-a-red-bull-themed-recruitment-phishing-campaign/
Researchers from The DFIR Report, in partnership with Proofpoint, have identified a new and resilient variant of the Interlock ransomware group’s remote access trojan (RAT). This new malware, a shift from the previously identified JavaScript-based [Interlock RAT](https://blog.sekoia.io/interlock-ransomware-evolving-under-the-radar/) (aka [NodeSnake](https://www.quorumcyber.com/wp-content/uploads/2025/06/20250416-Higher-Education-Sector-RAT.pdf)), uses PHP and is being used in a widespread campaign.
Hey folks,
Just published a write-up where I turned a blind XSS into Remote Code Execution , and the final step?
Injecting commands via Accept-Language header, parsed by a vulnerable PHP script.
No logs. No alert. Just clean shell access.
Would love to hear your thoughts or similar techniques you've seen!
🧠🛡️
[https://is4curity.medium.com/from-blind-xss-to-rce-when-headers-became-my-terminal-d137d2c808a3](https://is4curity.medium.com/from-blind-xss-to-rce-when-headers-became-my-terminal-d137d2c808a3)
I recently tested a language-learning site that used live frontend filtering to block HTML input (e.g., <img> <svg> tags were removed as you typed).
But by injecting the payload directly via browser console (without typing it), the input was submitted and stored.
Surprisingly, the XSS executed later on my own profile page — indicating stored execution from a DOM-based bypass.
I wrote a short write-up here:
[https://is4curity.medium.com/xss-before-submit-a-dom-based-execution-flaw-hidden-in-plain-sight-5633bdd686c9](https://is4curity.medium.com/xss-before-submit-a-dom-based-execution-flaw-hidden-in-plain-sight-5633bdd686c9)
enjoy
Just found [zorrochain.foundation](https://zorrochain.foundation), and it’s building a governance infrastructure that’s fully airgap-compatible.
No token, no speculation — just a system using entropy harvesting, offline ID, QR/USB syncing, and vault-based consensus.
No idea if it’s being audited, but conceptually it’s interesting for people thinking about zero-trust systems in collapse conditions.