Evaluating Large Language Models for Symbolic Security Protocol Analysis

Security protocol verification relies on formal tools such as ProVerif and OFMC. This study evaluates whether Large Language Models LLMs can perform comparable analysis. We test GPT and DeepSeek in chat and reasoning modes over three runs on 130 obfuscated AnB/AnBx protocols covering 388 security...

OpenSkillRisk: Benchmarking Agent Safety When Using Real-World Risky Third-Party Skills

LLM-based agents leverage third-party skills to extend their capabilities in open-world scenarios. However, third-party skills can introduce extra security vulnerabilities, as seemingly harmless skills can contain latent safety risks that only emerge during actual execution. In this work, we...

DARWIN: Evolving Jailbreak Adversary and Guardrail for LLM Safety Evaluation and Protection

Most existing LLM safety evaluation and defense methods follow a static formulation: jailbreak vulnerabilities are evaluated with fixed attack methods, and guardrails are trained on fixed malicious prompt datasets. However, real-world adversaries continuously evolve their capabilities and expand...

Defense against LLM Backdoors Using Critical Neuron Isolation Pruning

Large language models LLMs are vulnerable to backdoor attacks, where hidden triggers induce malicious outputs. Existing defenses generally fall into inference-time detection or training-time mitigation, but face two key limitations. First, they focus on fine-tuning-based backdoors e.g., PEFT...

CVE-2026-65012: Missing Authentication for Critical Function

InvokeAI before 6.13.7 contains an unauthenticated directory enumeration vulnerability in the GET /api/v2/models/scanfolder endpoint that accepts attacker-controlled scanpath parameters. Unauthenticated attackers can recursively enumerate arbitrary server filesystem directories and use HTTP...

OpenAI Models Escaped Containment and Hacked Hugging Face

The cybersecurity-focused models, including GPT-5.6 Sol, broke out of a testing sandbox, exploited a zero-day, and gained access to the open internet to pull off the attack...

Insertion of Sensitive Information Into Sent Data

Overview Affected versions of this package are vulnerable to Insertion of Sensitive Information Into Sent Data in the ToHook process. An attacker can access sensitive authorization credentials by making authorized API requests as a repository administrator. Remediation Upgrade...

Exposure of Private Personal Information to an Unauthorized Actor

Overview Affected versions of this package are vulnerable to Exposure of Private Personal Information to an Unauthorized Actor in the updateRepository function when a repository is transitioned from public to private via the API. An attacker can access sensitive repository metadata, such as...

Incorrect Authorization

Overview Affected versions of this package are vulnerable to Incorrect Authorization in the handling of API requests to /api/v1/teams/id when using a public-only access token. An attacker can retrieve private team repository metadata and private team activity feed entries by sending API requests...

Incorrect Authorization

Overview Affected versions of this package are vulnerable to Incorrect Authorization through changing a pull request's target branch without re-evaluating the official approval status on existing reviews. An attacker can bypass branch protection requirements and merge code into protected branches...

SUSE CVE-2026-64143

In the Linux kernel, the following vulnerability has been resolved: platform/x86: uniwill-laptop: Do not enable the charging limit even when forced It seems that on some older models 2020 the battery charging limit can permanently damage the battery. Prevent users from enabling this feature thru...

This Week in Spring - July 21st, 2026

Welcome to another installment of This Week in Spring! I'm writing this while enjoying a cup of coffee from the Caribbean on a little PTO. How're you doin'? Whenever I go on PTO, I like to sit by some body of water with a coffee and a laptop nearby. I alternate between plunges in the ocean and...

Find Before You Fine-Tune: A Diagnostic Study of Small LLMs for Cybersecurity QA

Large Language Models LLMs are increasingly fine-tuned for critical-domain Question-Answering QA, yet choosing which small model to adapt, before paying the cost of adaptation, remains difficult. Fine-tuning can improve domain alignment, but it may also erode prior knowledge, weaken...

Hugging Face Says Autonomous AI Agent System Breached Production Infrastructure

An AI-led cyberattack breached limited Hugging Face datasets and service credentials, while public models, Spaces and published packages showed no signs of tampering...

GHSA-33CG-GXV8-3P8G vLLM denial of service via prompt embeds on M-RoPE models

Summary Short summary of the problem. Make the impact and severity as clear as possible. For example: An unsafe deserialization vulnerability allows any unauthenticated user to execute arbitrary code on the server. Sending a pure prompt embeds payload in a /v1/completions request with a model usi...

vLLM denial of service via prompt embeds on M-RoPE models

Summary Short summary of the problem. Make the impact and severity as clear as possible. For example: An unsafe deserialization vulnerability allows any unauthenticated user to execute arbitrary code on the server. Sending a pure prompt embeds payload in a /v1/completions request with a model usi...

World's Largest AI Model Repository Hugging Face Breached by Autonomous AI Agent

In an ironic twist, open-source artificial intelligence AI platform Hugging Face revealed that it was the victim of a hack perpetrated by an autonomous AI agent system. The company said it detected and responded to the incident targeting its production infrastructure earlier last week. "We...

Broken Gates: Re-Evaluating Web Bot Defenses in the Age of LLM Agents

LLM-based browser agents are rapidly changing the threat landscape for web security. Unlike traditional automation frameworks that execute predefined scripts, these agents can autonomously navigate websites, reason about page content, and interact with web interfaces using natural-language...

Towards an Automated Test of LLM Security Knowledge

Large language models LLMs are increasingly used for a range of software, hardware and human-centered security tasks. Consequently, LLM performance on security tasks is an active area of measurement and research, often with a focus on identifying areas in which LLM security "knowledge'' may be...

CryptanalysisBench: Can LLMs Do Cryptanalysis?

Cryptanalysis - the task of finding attacks against cryptographic schemes - sits at the intersection of mathematical reasoning and cybersecurity, two areas where LLMs have advanced fastest. Cryptanalysis represents both a clean testbed for frontier reasoning as practical attacks can be...