Alphabet Index Mapping: Jailbreaking LLMs through Semantic Dissimilarity

Large Language Models LLMs have demonstrated remarkable capabilities, yet their susceptibility to adversarial attacks, particularly jailbreaking, poses significant safety and ethical concerns. While numerous jailbreak methods exist, many suffer from computational expense, high token usage, or...

AIRTBench: Measuring Autonomous AI Red Teaming Capabilities in Language Models

We introduce AIRTBench, an AI red teaming benchmark for evaluating language models' ability to autonomously discover and exploit Artificial Intelligence and Machine Learning AI/ML security vulnerabilities. The benchmark consists of 70 realistic black-box capture-the-flag CTF challenges from the...

A Halpha Metric for Identifying Dormant Black Holes in X-Ray Transients

Dormant black holes BHs in X-ray transients can be identified by the presence of broad Ha emission lines from quiescent accretion discs. Unfortunately, short-period cataclysmic variables CVs can also produce broad Ha lines, especially when viewed at high inclinations, and are thus a major source ...

Enhancing One-run Privacy Auditing with Quantile Regression-Based Membership Inference

Differential privacy DP auditing aims to provide empirical lower bounds on the privacy guarantees of DP mechanisms like DP-SGD. While some existing techniques require many training runs that are prohibitively costly, recent work introduces one-run auditing approaches that effectively audit DP-SGD...

Probe Before You Talk: Towards Black-Box Defense against Backdoor Unalignment for Large Language Models

Backdoor unalignment attacks against Large Language Models LLMs enable the stealthy compromise of safety alignment using a hidden trigger while evading normal safety auditing. These attacks pose significant threats to the applications of LLMs in the real-world Large Language Model as a Service...

Black-Box Privacy Attacks on Shared Representations in Multitask Learning

Multitask learning MTL has emerged as a powerful paradigm that leverages similarities among multiple learning tasks, each with insufficient samples to train a standalone model, to solve them simultaneously while minimizing data sharing across users and organizations. MTL typically accomplishes th...

IP Leakage Attacks Targeting LLM-Based Multi-Agent Systems

The rapid advancement of Large Language Models LLMs has led to the emergence of Multi-Agent Systems MAS to perform complex tasks through collaboration. However, the intricate nature of MAS, including their architecture and agent interactions, raises significant concerns regarding intellectual...

Astra Linux – Vulnerability found in Linux 6.1, Linux 6.12

In the Linux kernel, the following vulnerability has been resolved: schhfsc: made hfscqlennotify idempotent. hfscqlennotify itself is not idempotent, and it is not friendly to its callers, like fqcodeldequeue. We need to make it idempotent to ease the work of qdisctreereducebacklog callers: 1...

CVE-2025-45986

Blink routers BL-WR9000 V2.4.9 , BL-AC2100AZ3 V1.0.4, BL-X10AC8 v1.0.5 , BL-LTE300 v1.2.3, BL-F1200AT1 v1.0.0, BL-X26AC8 v1.2.8, BLAC450MAE4 v4.0.0 and BL-X26DA3 v1.2.7 werediscovered to contain a command injection vulnerability via the mac parameter in the bsSetMacBlack function...

Assessing the Resilience of Automotive Intrusion Detection Systems to Adversarial Manipulation

The security of modern vehicles has become increasingly important, with the controller area network CAN bus serving as a critical communication backbone for various Electronic Control Units ECUs. The absence of robust security measures in CAN, coupled with the increasing connectivity of vehicles,...

PT-2025-49372

Name of the Vulnerable Software and Affected Versions Linux kernel versions 3.14 through 6.18-rc5 Description A use-after-free UAF issue exists in the proc readdir de function within the Linux kernel. The problem occurs because rb erase is used to remove a proc directory entry pde from the...

Former Black Basta Members Use Microsoft Teams and Python Scripts in 2025 Attacks

Former members tied to the Black Basta ransomware operation have been observed sticking to their tried-and-tested approach of email bombing and Microsoft Teams phishing to establish persistent access to target networks. "Recently, attackers have introduced Python script execution alongside these...

GenBreak: Red Teaming Text-To-Image Generators Using Large Language Models

Text-to-image T2I models such as Stable Diffusion have advanced rapidly and are now widely used in content creation. However, these models can be misused to generate harmful content, including nudity or violence, posing significant safety risks. While most platforms employ content moderation...

BlackSuit Continues Social Engineering Attacks in Wake of Black Basta’s Internal Conflict

Executive Summary There has been a significant decrease in social engineering attacks linked to the Black Basta ransomware group since late December 2024. This lapse also included the leaked Black Basta chat logs in February 2025, indicating internal conflict within the group. Despite this, Rapid...

BlackSuit Continues Social Engineering Attacks in Wake of Black Basta’s Internal Conflict

Executive Summary There has been a significant decrease in social engineering attacks linked to the Black Basta ransomware group since late December 2024. This lapse also included the leaked Black Basta chat logs in February 2025, indicating internal conflict within the group. Despite this, Rapid...

HauntAttack: When Attack Follows Reasoning As a Shadow

Emerging Large Reasoning Models LRMs consistently excel in mathematical and reasoning tasks, showcasing exceptional capabilities. However, the enhancement of reasoning abilities and the exposure of their internal reasoning processes introduce new safety vulnerabilities. One intriguing concern is:...

Rewriting the Budget: a General Framework for Black-Box Attacks under Cost Asymmetry

Traditional decision-based black-box adversarial attacks on image classifiers aim to generate adversarial examples by slightly modifying input images while keeping the number of queries low, where each query involves sending an input to the model and observing its output. Most existing methods...

Differentially Private Explanations for Clusters

The dire need to protect sensitive data has led to various flavors of privacy definitions. Among these, Differential privacy DP is considered one of the most rigorous and secure notions of privacy, enabling data analysis while preserving the privacy of data contributors. One of the fundamental...

Prediction Inconsistency Helps Achieve Generalizable Detection of Adversarial Examples

Adversarial detection protects models from adversarial attacks by refusing suspicious test samples. However, current detection methods often suffer from weak generalization: their effectiveness tends to degrade significantly when applied to adversarially trained models rather than naturally train...

Tarallo: Evading Behavioral Malware Detectors in the Problem Space

Machine learning algorithms can effectively classify malware through dynamic behavior but are susceptible to adversarial attacks. Existing attacks, however, often fail to find an effective solution in both the feature and problem spaces. This issue arises from not addressing the intrinsic...