Kraken: Higher-Order EM Side-Channel Attacks on DNNs in near and Far Field

The multi-million dollar investment required for modern machine learning ML has made large ML models a prime target for theft. In response, the field of model stealing has emerged. Attacks based on physical side-channel information have shown that DNN model extraction is feasible, even on CUDA...

CVE-2025-62814

The CVE identifies a NULL pointer dereference in the Samsung Mobile Processor Exynos 1280, 2200, 1380, 1480, and 2400 series, specifically in ft_handle within load_fw_utc_vector(), which can cause a denial of service. Affected components are the Exynos mobile processors listed; the underlying cau...

CVE-2025-62817

An issue was discovered in Samsung Mobile Processor Exynos 1280, 2200, 1380, 1480, 2400, 1580, and 2500. A NULL pointer dereference of session-ncphdrbuf in pilotparsingncp causes a denial of service...

Incorrect Authorization

Overview openclaw is a 🦞 OpenClaw — Personal AI Assistant Affected versions of this package are vulnerable to Incorrect Authorization in the stop triggers and /models command. An attacker can disrupt active sessions and access sensitive model or authentication metadata by sending unauthorized...

OpenClaw has an unauthorized sender bypass in its stop triggers and /models command authorization

Summary Unauthorized senders could trigger two command paths without sender authorization checks: 1. stop-like natural-language abort triggers 2. /models command output Impact An unauthorized sender could disrupt active sessions and view model/auth metadata that should be authorization-gated. Fix...

GHSA-8M9V-XPGF-G99M OpenClaw has an unauthorized sender bypass in its stop triggers and /models command authorization

Summary Unauthorized senders could trigger two command paths without sender authorization checks: 1. stop-like natural-language abort triggers 2. /models command output Impact An unauthorized sender could disrupt active sessions and view model/auth metadata that should be authorization-gated. Fix...

LLM-Assisted Deanonymization

Turns out that LLMs are good at de-anonymization: We show that LLM agents can figure out who you are from your anonymous online posts. Across Hacker News, Reddit, LinkedIn, and anonymized interview transcripts, our method identifies users with high precision ­ and scales to tens of thousands of...

ZeroDayBench: Evaluating LLM Agents on Unseen Zero-Day Vulnerabilities for Cyberdefense

Large language models LLMs are increasingly being deployed as software engineering agents that autonomously contribute to repositories. A major benefit these agents present is their ability to find and patch security vulnerabilities in the codebases they oversee. To estimate the capability of...

TraceGuard: Process-Guided Firewall against Reasoning Backdoors in Large Language Models

The deployment of Large Reasoning Models LRMs in high-stakes decision-making pipelines has introduced a novel and opaque attack surface: reasoning backdoors. In these attacks, the model's intermediate Chain-of-Thought CoT is manipulated to provide a linguistically plausible but logically fallacio...

Can LLMs Hack Enterprise Networks? -- Replicated Computational Results (RCR) Report

This is the Replicated Computational Results RCR Report for the paper "Can LLMs Hack Enterprise Networks?" The paper empirically investigates the efficacy and effectiveness of different LLMs for penetration-testing enterprise networks, i.e., Microsoft Active Directory Assumed-Breach Simulations...

Jailbreaking Embodied LLMs Via Action-Level Manipulation

Embodied Large Language Models LLMs enable AI agents to interact with the physical world through natural language instructions and actions. However, beyond the language-level risks inherent to LLMs themselves, embodied LLMs with real-world actuation introduce a new vulnerability: instructions tha...

A Systematic Study of LLM-Based Architectures for Automated Patching

Large language models LLMs have shown promise for automated patching, but their effectiveness depends strongly on how they are integrated into patching systems. While prior work explores prompting strategies and individual agent designs, the field lacks a systematic comparison of patching...

VEcho: A Paradigm Shift from Vulnerability Verification to Proactive Discovery with Large Language Models

Static Application Security Testing SAST tools often suffer from high false positive rates, leading to alert fatigue that consumes valuable auditing resources. Recent efforts leveraging Large Language Models LLMs as filters offer limited improvements; however, these methods treat LLMs as passive,...

CVE-2026-1442 Unitree UPK files Hard-Coded Key

Since the encryption algorithm used to protect firmware updates is itself encrypted using key material available to an attacker or anyone paying attention, the firmware updates may be altered by an unauthorized user, and then trusted by a Unitree product, such as the Unitree Go2 and other models...

CVE-2026-24498

Exposure of Sensitive Information to an Unauthorized Actor vulnerability in EFM-Networks, Inc. IpTIME T5008, EFM-Networks, Inc. IpTIME AX2004M, EFM-Networks, Inc. IpTIME AX3000Q, EFM-Networks, Inc. IpTIME AX6000M allows Authentication Bypass.This issue affects ipTIME T5008: through 15.26.8; ipTIM...

Exploring Robust Intrusion Detection: A Benchmark Study of Feature Transferability in IoT Botnet Attack Detection

Cross-domain intrusion detection remains a critical challenge due to significant variability in network traffic characteristics and feature distributions across environments. This study evaluates the transferability of three widely used flow-based feature sets Argus, Zeek and CICFlowMeter across...

LLMs Generate Predictable Passwords

LLMs are bad at generating passwords: There are strong noticeable patterns among these 50 passwords that can be seen easily: All of the passwords start with a letter, usually uppercase G, almost always followed by the digit 7. Character choices are highly uneven ­ for example, L , 9, m, 2, $ and...

PT-2026-22151

Name of the Vulnerable Software and Affected Versions Flair versions 0.4.1 through latest Description The deserialization of untrusted data in the LanguageModel class can lead to arbitrary code execution when loading a malicious model. Recommendations Versions prior to 0.4.1 are not affected. At...

PT-2026-22213

Name of the Vulnerable Software and Affected Versions Manyfold versions prior to 0.133.1 Description Manyfold is a self-hosted web application for managing 3d models. A flaw exists in the get model method within the ModelFilesController lines 158-160 where models are loaded using Model.find...

Apple Security Update: visionOS 26.3.1

Apple recommends to install security update visionOS 26.3.1 on devices Apple Vision Pro all models...