Model Tampering Attacks and Detection
Research on understanding, performing, and defending against targeted modifications to LLM weights and behavior through model tampering
Model Tampering Attacks and Detection
Research on Large Language Models in Model Tampering Attacks and Detection
Beyond Inputs: Probing LLM Security Vulnerabilities
Revealing hidden capabilities through model tampering attacks
Hidden Threats in LLM Merging
How phishing models can compromise privacy in merged language models
Defending Against Code Poisoning Attacks
A lightweight detection method to protect neural code models
Stealing PII Through Model Merging
A novel security vulnerability in LLM integration processes
BadVision: The Backdoor Threat to Vision Language Models
How stealthy attacks can induce hallucinations in LVLMs
Securing LLMs Against Backdoor Attacks
New benchmark for evaluating LLM vulnerabilities
PoisonedParrot: The Subtle Threat to LLM Security
How data poisoning can trick LLMs into generating copyrighted content
Backdoor Vulnerabilities in Multi-modal AI
Exposing token-level security risks in image-text AI systems
Detecting Backdoor Threats in Outsourced AI Models
A novel cross-examination framework for identifying embedded backdoors
Hidden Threats in Large Language Models
New research reveals how LLMs can reproduce dangerous content verbatim
The Trojan in Your Model: LLM Security Alert
How malicious fine-tuning can weaponize language models
