AI red teaming agents change how LLMs get tested 2026-05-21 at 08:00 By Mirko Zorz Adversarial probing of LLMs has piled up a sprawling toolkit over the past three years. Attack techniques with names like Tree of Attacks with Pruning, Crescendo, and Skeleton Key sit alongside hundreds of prompt transforms and scoring methods across open-source […]
AI red teaming agents change how LLMs get tested Read More »
When your AI assistant has the keys to production 2026-05-20 at 09:34 By Sinisa Markovic Large language models in operational roles query telemetry, propose configuration changes, and in some deployments execute those changes against live infrastructure. Ticket drafting and alert summarization were the starting point. Vendors describe this work as autonomous remediation or self-healing infrastructure.
When your AI assistant has the keys to production Read More »
Earbud sensors can authenticate users by their heartbeat, study finds 2026-05-19 at 09:17 By Mirko Zorz Researchers built a continuous authentication system called AccLock that identifies a wearer by the tiny vibrations a heartbeat makes inside the ear canal. The signal comes from an accelerometer of the kind already sitting inside many wireless earbuds, so
Earbud sensors can authenticate users by their heartbeat, study finds Read More »
Public Instagram posts provide raw material for AI phishing campaigns 2026-05-19 at 09:17 By Sinisa Markovic A handful of public Instagram posts can give attackers enough material to generate convincing phishing emails with GenAI. Research from the University of Texas at Arlington and Louisiana State University showed how public social media activity can be turned
Public Instagram posts provide raw material for AI phishing campaigns Read More »
The AI backdoor your security stack is not built to see 2026-05-18 at 09:42 By Sinisa Markovic Enterprises deploying LLMs have spent the past two years building defenses around a reasonable assumption: malicious behavior leaves a trace in the input. Scan for suspicious tokens, filter unusual characters, watch for prompt injection patterns. New research from
The AI backdoor your security stack is not built to see Read More »
Deepfake detection is losing ground to generative models 2026-05-15 at 09:04 By Sinisa Markovic Deepfake detection has been built around a single question for close to a decade. Given a video or audio clip, is it real or synthetic? Commercial detectors analyze pixels, frequencies, and biometric signals to answer that question, and the best of
Deepfake detection is losing ground to generative models Read More »
Zombie linkages are keeping expired domains trusted for years 2026-05-15 at 08:24 By Sinisa Markovic Domains expire, get transferred, and return to the market every day. The systems connected to those domains can continue trusting the original owner long after control has changed. Researchers at USC and the University of Twente examined this problem in
Zombie linkages are keeping expired domains trusted for years Read More »
Vector embedding security gap exposes enterprise AI pipelines 2026-05-14 at 08:30 By Mirko Zorz Enterprise adoption of retrieval-augmented generation has moved sensitive corporate content into a new storage format that existing security tools cannot inspect. Companies deploying internal AI assistants convert documents into high-dimensional numerical vectors and ship them to embedding services and vector databases
Vector embedding security gap exposes enterprise AI pipelines Read More »
The hidden smart fridge risks that emerge years after purchase 2026-05-12 at 09:28 By Mirko Zorz Household refrigerators are built to last more than a decade. The software, cloud services, and mobile apps that control them are not. A new analysis from Erik Buchmann at Leipzig University maps what happens when those two timelines collide,
The hidden smart fridge risks that emerge years after purchase Read More »
One keypress is all it takes to compromise four AI coding tools 2026-05-08 at 01:14 By Mirko Zorz Developers clone unfamiliar repositories all the time. Open-source projects, work from teammates, sample code from a tutorial, a library someone recommended on a forum. The convention is old and reasonable: you look at what’s inside before you
One keypress is all it takes to compromise four AI coding tools Read More »
Can your coding style predict whether your code is vulnerable? 2026-05-05 at 13:21 By Sinisa Markovic Developers leave fingerprints in the code they write. Naming choices, indentation patterns, preferred APIs, and the way someone structures a loop or handles a pointer all carry traces of individual habit. Researchers have used these stylistic signals for years
Can your coding style predict whether your code is vulnerable? Read More »
What researchers learned about building an LLM security workflow 2026-05-04 at 09:46 By Sinisa Markovic Security operations centers are running into the same wall everywhere. Detection tools generate more alerts than analysts can work through, and the early stages of any investigation involve pulling together logs from several sources to decide whether something is worth
What researchers learned about building an LLM security workflow Read More »
Researchers develop tool to expose GPS signal spoofing in transit networks 2026-04-30 at 15:31 By Anamarija Pogorelec The Oak Ridge National Laboratory (ORNL) has developed a portable detector that identifies GPS spoofing in real time, including during motion, to help protect transportation systems. Spoofing involves transmitting counterfeit signals that imitate authentic GPS transmissions and produce
Researchers develop tool to expose GPS signal spoofing in transit networks Read More »
The AI criminal mastermind is already hiring on gig platforms 2026-04-27 at 10:30 By Mirko Zorz Labor-hire platforms let anyone with a credit card post a task and pay a stranger to complete it. The RentAHuman platform extends that model to AI agents through a Model Context Protocol server, allowing an agent to post gigs
The AI criminal mastermind is already hiring on gig platforms Read More »
Indirect prompt injection is taking hold in the wild 2026-04-24 at 23:26 By Zeljka Zorz The open web is slowly but surely filling up with “traps” designed for LLM-powered AI agents. The technique, known as indirect prompt injection (IPI), involves hiding (more or less) covert instructions inside ordinary web pages, waiting for an AI agent
Indirect prompt injection is taking hold in the wild Read More »
A study of 1,000 Android apps finds a privacy policy logging gap 2026-04-24 at 08:22 By Anamarija Pogorelec Android developers write log statements for the same reasons they always have: debugging crashes, tracing performance issues, and understanding how features behave in production. Legal and privacy teams, working from templates and regulatory checklists, draft policies describing
A study of 1,000 Android apps finds a privacy policy logging gap Read More »
GDPR works, but only where someone enforces it 2026-04-23 at 07:32 By Sinisa Markovic A new measurement study of web tracking across ten countries offers a reality check for anyone working on privacy compliance. Researchers crawled the same set of globally popular websites from virtual machines located in Australia, Brazil, Canada, Germany, India, Singapore, South
GDPR works, but only where someone enforces it Read More »
Apple Intelligence flaw kept stolen tokens reusable on another device 2026-04-22 at 10:09 By Sinisa Markovic Apple claims that Apple Intelligence, a GenAI service provided on its operating systems, is designed with an extra focus on user security and privacy through a two-stage authentication and authorization system using anonymous access tokens. However, researchers from The
Apple Intelligence flaw kept stolen tokens reusable on another device Read More »
Researchers build an encrypted routing layer for private AI inference 2026-04-21 at 07:31 By Sinisa Markovic Organizations in healthcare, finance, and other sensitive industries want to use large AI models without exposing private data to the cloud servers running those models. A cryptographic technique called Secure Multi-Party Computation (MPC) makes this possible. It splits data
Researchers build an encrypted routing layer for private AI inference Read More »
Half of the 6 Million Internet-Facing FTP Servers Lack Encryption 2026-04-20 at 14:37 By Ionut Arghire The continued use of the half-century-old protocol exposes enterprises and end users to various types of attacks. The post Half of the 6 Million Internet-Facing FTP Servers Lack Encryption appeared first on SecurityWeek. This article is an excerpt from
Half of the 6 Million Internet-Facing FTP Servers Lack Encryption Read More »