AI Threat AlertSAGE: Similarity-Based Cleaning of Poisoned Training Data from Verified Examples
clean and poisoned ones, i.e., each example verified either to be clean or poisoned through inspection by a forensic expert. The challenge is then to detect poisons based
Knowledge Poisoning Attacks on Medical Multi-Modal Retrieval-Augmented Generation
injected with, adversarial knowledge, which can perturb model outputs and undermine system reliability. To investigate this risk, prior studies have explored knowledge poisoning attacks in medical RAG systems. Nevertheless, most
TokenSwap: Backdoor Attack on the Compositional Understanding of Large Vision-Language Models
corresponding textual answers. However, the poisoned samples exhibit only subtle differences from the original ones, making it challenging for the model to learn the backdoor behavior. To address this, TokenSwap
The Like Trap: Multi-Stage Poisoning against Agents in Similarity-based Recommendation Systems
With recent advancements in large language models (LLMs) and LLM-based agents, these agents are becoming increasingly autonomous and gaining broader access to act on users' behalf on the internet
A Layered Security Framework Against Prompt Injection in RAG-Based Chatbots
cannot prevent malicious payloads from reaching the model. Consequently, retrieval-augmented generation (RAG) chatbots remain vulnerable to indirect injection, where a poisoned knowledge-base document compromises every user whose query
Combating Knowledge Corruption in Agent Systems: A Byzantine-Tolerant Secure Collaborative RAG Framework
address the hallucination issues in large language models, it also introduces new vulnerabilities to knowledge corruption attacks. Adversaries exploit these vulnerabilities by poisoning documents provided by RAG system to manipulate
Backdoor Threats in Variational Quantum Circuits: Taxonomy, Attacks, and Defenses
survey of backdoor attacks in VQCs, covering data-poisoning, compiler-level, and quantum-native mechanisms. We formalize key terminology and threat models, and review existing attack strategies along with their
IU: Imperceptible Universal Backdoor Attack
simultaneously controls all target classes with minimal poisoning while preserving stealth. Our key idea is to leverage graph convolutional networks (GCNs) to model inter-class relationships and generate class-specific
DropVLA: An Action-Level Backdoor Attack on Vision-Language-Action Models
tuning. On OpenVLA-7B evaluated with LIBERO, vision-only poisoning achieves 98.67%-99.83% attack success rate (ASR) with only 0.31% poisoned episodes while preserving 98.50%-99.17% clean-task retention
Your Agent, Their Asset: A Real-World Safety Analysis of OpenClaw
live OpenClaw instance across four backbone models (Claude Sonnet 4.5, Opus 4.6, Gemini 3.1 Pro, and GPT-5.4). The results show that poisoning any single CIK dimension increases the average
ToxicTextCLIP: Text-Based Poisoning and Backdoor Attacks on CLIP Pre-training
Contrastive Language-Image Pretraining (CLIP) model has significantly advanced vision-language modeling by aligning image-text pairs from large-scale web data through self-supervised contrastive learning. Yet, its reliance
KEPo: Knowledge Evolution Poison on Graph-based Retrieval-Augmented Generation
timeliness and accuracy of Large Language Model (LLM) generations. However, this reliance on external data introduces new attack surfaces. Attackers can inject poisoned texts into databases to manipulate LLMs into
When Modalities Fail to Tango: Conformal Backdoor Detection in Multimodal Contrastive Learning
tasks critically depend on pre-trained MCL models. Existing detection-based defenses predominantly rely on the CLIPScore metric, under the assumption that poisoned pairs exhibit lower semantic similarity between
MIRAGE: Misleading Retrieval-Augmented Generation via Black-box and Query-agnostic Poisoning Attacks
proposing MIRAGE, a novel multi-stage poisoning pipeline designed for strict black-box and query-agnostic environments. Operating on surrogate model feedback, MIRAGE functions as an automated optimization framework that
SPOILER: TEE-Shielded DNN Partitioning of On-Device Secure Inference with Poison Learning
Deploying deep neural networks (DNNs) on edge devices exposes valuable intellectual property to model-stealing attacks. While TEE-shielded DNN partitioning (TSDP) mitigates this by isolating sensitive computations, existing paradigms
CryptoGuard: Lightweight Hybrid Detection and Response to Host-based Cryptojackers in Linux Cloud Environments
phase process, leveraging deep learning models to identify suspicious activity with high precision. To counter evasion techniques such as entry point poisoning and PID manipulation, CryptoGuard integrates targeted remediation mechanisms
DF-LoGiT: Data-Free Logic-Gated Backdoor Attacks in Vision Transformers
backdoor attacks largely rely on poisoned-data training, while prior data-free attempts typically require synthetic-data fine-tuning or extra model components. This paper introduces Data-Free Logic-Gated
Poisoning the Neighborhood: A Topology-Aware Topic-Level Attack on Large-Scale HNSW-based RAG Systems
knowledge staleness in large language models using external knowledge. However, ingestion from public and third-party sources exposes RAG systems to poisoning attacks. Existing attacks are designed and evaluated under
Diversity Matters: Distributional Feature Coverage Sample Selection for Data-Efficient Backdoor Attacks
compromise training data so that a model retains clean accuracy but predicts an attacker-chosen target on triggered inputs. At very low poisoning rates, only a few samples convey
When Backdoors Go Beyond Triggers: Semantic Drift in Diffusion Models Under Encoder Attacks
attacks on text-to-image (T2I) models primarily measure trigger activation and visual fidelity. We challenge this paradigm, demonstrating that encoder-side poisoning induces persistent, trigger-free semantic corruption that