BESA: Boosting Encoder Stealing Attack with Perturbation Recovery

To boost the encoder stealing attack under the perturbation-based defense that hinders the attack performance, we propose a boosting encoder stealing attack with perturbation recovery named BESA. It aims to overcome perturbation-based defenses. The core of BESA consists of two modules: perturbati...

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...

Watermarking Degrades Alignment in Language Models: Analysis and Mitigation

Watermarking techniques for large language models LLMs can significantly impact output quality, yet their effects on truthfulness, safety, and helpfulness remain critically underexamined. This paper presents a systematic analysis of how two popular watermarking approaches-Gumbel and KGW-affect...

Razer Synapse 4 安全漏洞

Razer Synapse 4 is an application from the American company Razer, Inc. cloud-based unified hardware configuration tool. A security vulnerability exists in Razer Synapse 4 4.0.86.2502180127 and prior versions, which stems from a COM interface vulnerability that could lead to local elevation of...

ATAG: AI-Agent Application Threat Assessment with Attack Graphs

Evaluating the security of multi-agent systems MASs powered by large language models LLMs is challenging, primarily because of the systems' complex internal dynamics and the evolving nature of LLM vulnerabilities. Traditional attack graph AG methods often lack the specific capabilities to model...

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...

BadReward: Clean-Label Poisoning of Reward Models in Text-To-Image RLHF

Reinforcement Learning from Human Feedback RLHF is crucial for aligning text-to-image T2I models with human preferences. However, RLHF's feedback mechanism also opens new pathways for adversaries. This paper demonstrates the feasibility of hijacking T2I models by poisoning a small fraction of...

Privacy Leaks by Adversaries: Adversarial Iterations for Membership Inference Attack

Membership inference attack MIA has become one of the most widely used and effective methods for evaluating the privacy risks of machine learning models. These attacks aim to determine whether a specific sample is part of the model's training set by analyzing the model's output. While traditional...

MISLEADER: Defending against Model Extraction with Ensembles of Distilled Models

Model extraction attacks aim to replicate the functionality of a black-box model through query access, threatening the intellectual property IP of machine-learning-as-a-service MLaaS providers. Defending against such attacks is challenging, as it must balance efficiency, robustness, and utility...

COALESCE: Economic and Security Dynamics of Skill-Based Task Outsourcing among Team of Autonomous LLM Agents

The meteoric rise and proliferation of autonomous Large Language Model LLM agents promise significant capabilities across various domains. However, their deployment is increasingly constrained by substantial computational demands, specifically for Graphics Processing Unit GPU resources. This pape...

The vulnerability of Junos OS and Junos OS Evolved routers from the MX, MX304, EX9200, and PTX Series series lies in the insufficient handling of exceptional states, allowing attackers to trigger service interruptions.

The vulnerability of Junos OS and Junos OS Evolved routers from the MX, MX304, EX9200, and PTX Series series lies in the insufficient handling of exceptional states. Exploiting this vulnerability can allow a malicious actor to cause service interruptions...

Dot 跨站脚本漏洞

Dot is a text-to-speech, RAG and LLM tool by alexpinel individual developers. A cross-site scripting vulnerability exists in Dot 0.9.3 and earlier versions, which stems from user input and LLM output being appended to the DOM using innerHTML, which could lead to cross-site scripting and command...

CSVAR: Enhancing Visual Privacy in Federated Learning Via Adaptive Shuffling against Overfitting

Although federated learning preserves training data within local privacy domains, the aggregated model parameters may still reveal private characteristics. This vulnerability stems from clients' limited training data, which predisposes models to overfitting. Such overfitting enables models to...

Formal Security Analysis of SPV Clients Versus Home-Based Full Nodes in Bitcoin-Derived Systems

This paper presents a mathematically rigorous formal analysis of Simplified Payment Verification SPV clients, as specified in Section 8 of the original Bitcoin white paper, versus non-mining full nodes operated by home users. It defines security as resistance to divergence from global consensus a...

Sitecom WLX-2006 Wall Mount Range Extender N300 安全漏洞

The Sitecom WLX-2006 Wall Mount Range Extender N300 is a wall mount range extender from Sitecom. A security vulnerability exists in the Sitecom WLX-2006 Wall Mount Range Extender N300 v1.5 and earlier versions, which stems from a mishandling of a language cookie that could result in local file...

Black-Box Crypto Is Useless for Pseudorandom Codes

A pseudorandom code is a keyed error-correction scheme with the property that any polynomial number of encodings appear random to any computationally bounded adversary. We show that the pseudorandomness of any code tolerating a constant rate of random errors cannot be based on black-box reduction...

ReGA: Representation-Guided Abstraction for Model-Based Safeguarding of LLMs

Large Language Models LLMs have achieved significant success in various tasks, yet concerns about their safety and security have emerged. In particular, they pose risks in generating harmful content and vulnerability to jailbreaking attacks. To analyze and monitor machine learning models,...

Developing a Risk Identification Framework for Foundation Model Uses

As foundation models grow in both popularity and capability, researchers have uncovered a variety of ways that the models can pose a risk to the model's owner, user, or others. Despite the efforts of measuring these risks via benchmarks and cataloging them in AI risk taxonomies, there is little...

Nearly-Linear Time Private Hypothesis Selection with the Optimal Approximation Factor

Estimating the density of a distribution from its samples is a fundamental problem in statistics. Hypothesis selection addresses the setting where, in addition to a sample set, we are given $n$ candidate distributions -- referred to as hypotheses -- and the goal is to determine which one best...

Con Instruction: Universal Jailbreaking of Multimodal Large Language Models Via Non-Textual Modalities

Existing attacks against multimodal language models MLLMs primarily communicate instructions through text accompanied by adversarial images. In contrast, we exploit the capabilities of MLLMs to interpret non-textual instructions, specifically, adversarial images or audio generated by our novel...