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...

LLM Jailbreak Oracle

As large language models LLMs become increasingly deployed in safety-critical applications, the lack of systematic methods to assess their vulnerability to jailbreak attacks presents a critical security gap. We introduce the jailbreak oracle problem: given a model, prompt, and decoding strategy,...

AlphaSteer: Learning Refusal Steering with Principled Null-Space Constraint

As LLMs are increasingly deployed in real-world applications, ensuring their ability to refuse malicious prompts, especially jailbreak attacks, is essential for safe and reliable use. Recently, activation steering has emerged as an effective approach for enhancing LLM safety by adding a refusal...

From Threat to Tool: Leveraging Refusal-Aware Injection Attacks for Safety Alignment

Safely aligning large language models LLMs often demands extensive human-labeled preference data, a process that's both costly and time-consuming. While synthetic data offers a promising alternative, current methods frequently rely on complex iterative prompting or auxiliary models. To address...

Benchmarking Misuse Mitigation against Covert Adversaries

Existing language model safety evaluations focus on overt attacks and low-stakes tasks. Realistic attackers can subvert current safeguards by requesting help on small, benign-seeming tasks across many independent queries. Because individual queries do not appear harmful, the attack is hard to...

Mind the Gap: a Practical Attack on GGUF Quantization

With the increasing size of frontier LLMs, post-training quantization has become the standard for memory-efficient deployment. Recent work has shown that basic rounding-based quantization schemes pose security risks, as they can be exploited to inject malicious behaviors into quantized models tha...

Security update for wxWidgets-3_2

This update for wxWidgets-32 fixes the following issues: CVE-2024-58249: Fixed crash when connection is refused in wxWebRequestCURL bsc1239902 Patch Instructions: To install this SUSE update use the SUSE recommended installation methods like YaST onlineupdate or "zypper patch". Alternatively you...

TokenProber: Jailbreaking Text-To-Image Models Via Fine-Grained Word Impact Analysis

Text-to-image T2I models have significantly advanced in producing high-quality images. However, such models have the ability to generate images containing not-safe-for-work NSFW content, such as pornography, violence, political content, and discrimination. To mitigate the risk of generating NSFW...

Steering the CensorShip: Uncovering Representation Vectors for LLM "Thought" Control

Large language models LLMs have transformed the way we access information. These models are often tuned to refuse to comply with requests that are considered harmful and to produce responses that better align with the preferences of those who control the models. To understand how this "censorship...

AiXamine: Simplified LLM Safety and Security

Evaluating Large Language Models LLMs for safety and security remains a complex task, often requiring users to navigate a fragmented landscape of ad hoc benchmarks, datasets, metrics, and reporting formats. To address this challenge, we present aiXamine, a comprehensive black-box evaluation...

CVE-2024-58249

In wxWidgets before 3.2.7, a crash can be triggered in wxWidgets apps when connections are refused in wxWebRequestCURL...

PT-2025-16788 · Wxwidgets +2 · Wxwidgets +2

Name of the Vulnerable Software and Affected Versions: wxWidgets versions prior to 3.2.7 Description: A crash can be triggered in wxWidgets apps when connections are refused in wxWebRequestCURL. Recommendations: For versions prior to 3.2.7, update to version 3.2.7 or later to resolve the issue...

CVE-2024-58249

In wxWidgets before 3.2.7, a crash can be triggered in wxWidgets apps when connections are refused in wxWebRequestCURL...

Astra Linux – Vulnerability in Linux 6.1

In the Linux kernel, the following vulnerability has been resolved: remoteproc: mediatek: Ensure that the IPI buffer fits within the L2TCM. The location of the IPI buffer is determined from the firmware that we load into the System Companion Processor. It’s not guaranteed that both the SRAM size...

WordPress Contact Form 7 Telegram plugin <= 0.8.5 - Missing Authorization to Authenticated (Subscriber+) Subscription Approve/Pause/Refuse vulnerability

Missing Authorization to Authenticated Subscriber+ Subscription Approve/Pause/Refuse vulnerability discovered by István Márton in WordPress Plugin Contact Form 7 Telegram versions = 0.8.5...

SUSE CVE-2024-36965

In the Linux kernel, the following vulnerability has been resolved: remoteproc: mediatek: Make sure IPI buffer fits in L2TCM The IPI buffer location is read from the firmware that we load to the System Companion Processor, and it's not granted that both the SRAM L2TCM size that is defined in the...

CVE-2024-4425

The access control in CemiPark software stores integration e.g. FTP or SIP credentials in plain-text. An attacker who gained unauthorized access to the device can retrieve clear text passwords used by the system.This issue affects CemiPark software: 4.5, 4.7, 5.03 and potentially others. The vend...

CVE-2024-4423

The access control in CemiPark software does not properly validate user-entered data, which allows the authentication bypass. An attacker who has network access to the login panel can log in with administrator rights to the application.This issue affects CemiPark software: 4.5, 4.7, 5.03 and...

CVE-2024-4425

CVE-2024-4425 affects CemiPark software (versions 4.5, 4.7, 5.03 and potentially others). The root cause is improper handling of credentials, with integration credentials (e.g., FTP or SIP) stored in plain-text. An attacker who gains unauthorized access to the device can retrieve clear-text passw...

CVE-2024-4424

CVE-2024-4424 affects CemiPark software (versions 4.5, 4.7, 5.03 and potentially others) where input data is not properly validated, enabling stored cross-site scripting (XSS). The vulnerability arises from insufficient validation of user-entered data in the access control/data entry pathways, al...