ShadowMem

ShadowMem: Safeguarding LLM Agents against Long-Horizon Threats via Shadow Memory This repository contains the official code release for the paper Safeguarding LLM Agents against Long-Horizon Threats via Shadow Memory. ShadowMem is a defensive framework that maintains a dedicated, safety-focused...

AI and the Automobile — Updated!

Open Arsenal AI and the Autombile Specifications The AI and the Automobile collection brings together multiple specifications that define how AI integrates into modern vehicle systems. Across the set, it treats the automobile as a software-defined platform where intelligence is distributed across...

Securing Autonomous Vehicle Systems Via Twin-Aware Federated Reinforcement Learning

Federated reinforcement learning FRL is crucial for enabling collaborative learning across multiple agents without sharing raw data, thereby enhancing privacy and scalability in the decision-making process within dynamic vehicular environments. However, poisoning attacks pose a significant threat...

LLM Agent Safety, Multi-Turn Red-Teaming, Jailbreak Benchmarks, Adversarial Robustness, Safety-Critical Systems

Large language model LLM agents are increasingly proposed as supervisory components for safety-critical systems, yet their robustness under sustained, adaptive adversarial pressure remains poorly characterized. We present NRT-Bench, a benchmark for multi-turn red-teaming of LLM agents acting as...

Anywhere, Any-Stymie: Remote Activation of Trojan Malware on LiDAR with Modulated Signals

LiDAR sensors are widely deployed in autonomous systems for 3D perception and safety-critical decision-making. We identify a previously unexplored attack surface in which dormant malware embedded in the LiDAR sensing pipeline remains inactive during normal operation and can be externally triggere...

Large Language Models As Explainable Cyberattack Detectors for Energy Industrial Control Systems

In modern energy systems, industrial control systems ICS and power-system SCADA require intrusion detection that is not only accurate but also auditable by operators. The ICS intrusion-detection landscape is currently dominated by established supervised detectors. In this paper, we study whether ...

PT-2025-51840

Name of the Vulnerable Software and Affected Versions Netun Solutions HelpFlash IoT version v18 178 221102 ASCII PRO 1R5 50 Description The over-the-air OTA firmware update process in the software does not properly authenticate update servers or validate firmware signatures, and relies on...

SAFEx: Analyzing Vulnerabilities of MoE-Based LLMs Via Stable Safety-Critical Expert Identification

Large language models based on Mixture-of-Experts have achieved substantial gains in efficiency and scalability, yet their architectural uniqueness introduces underexplored safety alignment challenges. Existing safety alignment strategies, predominantly designed for dense models, are ill-suited t...

Secure Distributed Learning for CAVs: Defending against Gradient Leakage with Leveled Homomorphic Encryption

Federated Learning FL enables collaborative model training across distributed clients without sharing raw data, making it a promising approach for privacy-preserving machine learning in domains like Connected and Autonomous Vehicles CAVs. However, recent studies have shown that exchanged model...

A Survey of Learning-Based Intrusion Detection Systems for In-Vehicle Network

Connected and Autonomous Vehicles CAVs enhance mobility but face cybersecurity threats, particularly through the insecure Controller Area Network CAN bus. Cyberattacks can have devastating consequences in connected vehicles, including the loss of control over critical systems, necessitating robus...

Learning from the Good Ones: Risk Profiling-Based Defenses against Evasion Attacks on DNNs

Safety-critical applications such as healthcare and autonomous vehicles use deep neural networks DNN to make predictions and infer decisions. DNNs are susceptible to evasion attacks, where an adversary crafts a malicious data instance to trick the DNN into making wrong decisions at inference time...

High-Severity Flaws Uncovered in Bosch Thermostats and Smart Nutrunners

Multiple security vulnerabilities have been disclosed in Bosch BCC100 thermostats and Rexroth NXA015S-36V-B smart nutrunners that, if successfully exploited, could allow attackers to execute arbitrary code on affected systems. Romanian cybersecurity firm Bitdefender, which discovered the flaw in...

Vulnerability disclosure in aviation

We joined Boeing and United Airlines on a panel recently at the RSA Conference to talk about vulnerability disclosure in the aviation world. The engagement we are now seeing between researchers and industry is a powerful force for positive change. Hopefully this will start to reduce the number of...

DEF CON 30. Hacking EFBs. Engine Performance

At DEF CON 30 this year we demonstrated some vulnerabilities in electronic flight bags and the potential impact on flight safety. There’s plenty more detail of EFB security issues here. As part of the Aerospace Village at DEF CON 30, we invited people to fly our flight sim under instruction from...

Bluetooth + Electrical switchgear

The ongoing rapid growth of Industrial IoT IIoT across all business sectors continues to bring to focus the discrepancies that exist between the approaches to safety and cyber-security on safety critical sites. Safety has been culturally ingrained into all aspects of industrial site operations fo...

CVE-2022-2105

Client-side JavaScript controls may be bypassed to change user credentials and permissions without authentication, including a “root” user level meant only for the vendor. Web server root level access allows for changing of safety critical parameters...

CVE-2022-2105

Client-side JavaScript controls may be bypassed to change user credentials and permissions without authentication, including a “root” user level meant only for the vendor. Web server root level access allows for changing of safety critical parameters...

Authentication flaw

Client-side JavaScript controls may be bypassed to change user credentials and permissions without authentication, including a “root” user level meant only for the vendor. Web server root level access allows for changing of safety critical parameters...

CVE-2022-2105 Secheron SEPCOS Control and Protection Relay

Client-side JavaScript controls may be bypassed to change user credentials and permissions without authentication, including a “root” user level meant only for the vendor. Web server root level access allows for changing of safety critical parameters...

PT-2022-3518

Name of the Vulnerable Software and Affected Versions Secheron SEPCOS Single Package relay control and protection software affected versions not specified Description The issue is related to the incorrect implementation of a sequence of actions in the software. This could allow a remote attacker ...