271
Topics /
Security & Privacy
The most popular blog posts about security & privacy on Hacker News. 717 posts.
272
Brazilian Court Sanctions Lawyers for Prompt Injection in Court Filing
Two lawyers hid a white-on-white instruction inside a court filing telling the AI to go easy on their case. The court's AI caught it. The fine was $16,000.
273
Eye of Sauron: Long-Range Hidden Spy Camera Detection (2024)
Qibo Zhang and Daibo Liu, Hunan University; Xinyu Zhang, University of California San Diego; Zhichao Cao, Michigan State University; Fanzi Zeng, Hongbo Jiang, and Wenqiang Jin, Hunan University
274
This year’s insane timeline of hacks
An archive of this year’s insane timeline of hacks few people are talking about
275
Don't trust AI agents
AI agents need a security model that assumes things will go wrong. The right response isn't better permission checks, it's architecture that makes trust unnecessary.
276
Inside ZCode: Silently uploading your Git history to the cloud
ZCode silently uploads full Git history to the cloud; this post provides a block rule and source review — checkpoint claims fall apart against the code.
277
GPT-6 Astra covertly underperforms on safety research
Research on aligning AI with human values and intent, and reports documenting model failures.
278
Humans missed 1 in 3 threats approving AI agent commands across 40k game runs
Results from AI agent permission game: which attacks beat human reviewers, and which safe commands got blocked instead.
279
Azure's Weakest Link – Five Full Cross-Tenant Compromises
In my previous blog posts, Azure’s Weakest Link? and Azure’s Weakest Link - Full Cross-Tenant Compromise, I gave an overview of the severely insecure architecture behind API Connections in Azure, a part of Azure Logic Apps, and an instance of a full cross-tenant compromise using these inherent flaws. Now I am back…
280
Concerning behavior from a popular open-weight AI model
Discover how untrusted open-weight AI models can potentially leak data, even when self-hosted. Learn the security risks and how to protect your enterprise.
281
Why aren't AI companies being charged for hacking?
AI companies claim their hacks were unpredictable accidents — but they follow foreseeably from deliberate design choices, and that establishes criminal intent.
282
The Hugging Face incident and the road ahead
OpenAI shares findings from the Hugging Face security incident and the steps we’re taking to strengthen AI model security, monitoring, and alignment.
283
Top downloaded skill in ClawHub contains malware
The same capabilities that make OpenClaw a groundbreaking tool also make it an urgent security risk. This blog contains confirmed examples of agent skills being used as malware vectors, and advice on how to protect yourself if you're experimenting with them.
284
Meta's ships facial recognition on smart glasses
Stella is the companion app for Meta's smart glasses. Inspecting version 273.0.0.21 of the Android build (com.facebook.stella), I found the entire computational and storage stack for on-device facial recognition: three face models, a local database schema, a cosine-similarity vector index dimensioned to
285
OpenAI Adopts Google's SynthID Watermark for AI Images with Verification Tool
OpenAI advances AI content provenance with Content Credentials, SynthID, and a verification tool to help people identify and trust AI-generated media.
286
Chained PHP2Shell exploit roots 25 servers, including US Government servers
Custom PHP2Shell Exploit Chained with Kernel LPE to Root Versions of Palo Alto Networks PAN-OS Firewalls in the Wild
287
We got admin access to Baseten's production GitHub
We gave Strix a domain. In 25 minutes, it found a live token with admin access to Baseten's product, deployment, and CLI repos in a public Docker image.
288
My audio interface has SSH enabled by default
I bought a Rodecaster Duo to solve some audio woes, and found that it has ssh enabled by default. I captured the firmware update process and created custom firmware to enable password authentication for ssh.
289
Bucketsquatting is finally dead
For a decade, I have been working with AWS and third-party security teams to resolve bucketsquatting / bucketsniping issues in AWS S3. Finally, I am happy to say AWS now has a solution to the problem, and it changes the way you should name your buckets.
290
291
Android Developer Verification
News and insights on the Android platform, developer tools, and events.
292
Assessing Claude Mythos Preview's cybersecurity capabilities
Technical details on how we tested Claude Mythos Preview, a model strikingly capable at computer security tasks, and what we found.
293
EY Canada published a cybersecurity report and most citations were hallucinated
An investigation into fabricated citations and broken references in EY's "Points of Attack" cybersecurity report on loyalty system fraud.
294
German police name alleged leaders of GandCrab and REvil ransomware groups
An elusive hacker who went by the handle “ UNKN ” and ran the early Russian ransomware groups GandCrab and REvil now has a name and a face. Authorities in Germany say 31-year-old Russian Daniil Maksimovich Shchukin headed both cybercrime gangs and helped carry out at least 130 acts of computer sabotage and extortion…
295
I could've accessed 17T Microsoft records
How one unchecked login token put 17 trillion rows in a Microsoft internal analytics service within reach.
296
New information extracted from Snowden PDFs through metadata version analysis
We discovered that entire sections describing domestic U.S. intelligence facilities were deliberately removed from two published documents, while equivalent foreign facilities remained visible. The evidence exists in an unexpected place - the PDF metadata of documents published by The Intercept in 2016, and by The…
297
ChatGPT for Google Sheets exfiltrates workbooks
ChatGPT for Google Sheets is vulnerable to data exfiltration and phishing overlay attacks that affect workbooks across the victim’s account after an indirect prompt injection in a single sheet.
298
Stripe is friendly to “friendly fraud”
Update: this post generated some interesting discussion on Hacker News. Friendly fraud is the laundered name for something that the payment system is not really able to prevent. Even though I’m pretty sure they can do way better. Particularly big and sophisticated payment providers like Stripe, with a mountain of…
299
The RCE that AMD wouldn't fix
Following backlash, AMD agreed to fix an MITM RCE I discovered in Ryzen Master. All in all, it took them over four months to roll out the fix.
300
DNS-Persist-01: A New Model for DNS-Based Challenge Validation
When you request a certificate from Let’s Encrypt, our servers validate that you control the hostnames in that certificate using ACME challenges. For subscribers who need wildcard certificates or who prefer not to expose infrastructure to the public Internet, the DNS-01 challenge type has long been the only choice…