Alert button
Picture for Zhen Xiang

Zhen Xiang

Alert button

ArtPrompt: ASCII Art-based Jailbreak Attacks against Aligned LLMs

Feb 22, 2024
Fengqing Jiang, Zhangchen Xu, Luyao Niu, Zhen Xiang, Bhaskar Ramasubramanian, Bo Li, Radha Poovendran

Viaarxiv icon

BadChain: Backdoor Chain-of-Thought Prompting for Large Language Models

Jan 20, 2024
Zhen Xiang, Fengqing Jiang, Zidi Xiong, Bhaskar Ramasubramanian, Radha Poovendran, Bo Li

Viaarxiv icon

CBD: A Certified Backdoor Detector Based on Local Dominant Probability

Oct 26, 2023
Zhen Xiang, Zidi Xiong, Bo Li

Viaarxiv icon

Backdoor Mitigation by Correcting the Distribution of Neural Activations

Aug 18, 2023
Xi Li, Zhen Xiang, David J. Miller, George Kesidis

Figure 1 for Backdoor Mitigation by Correcting the Distribution of Neural Activations
Figure 2 for Backdoor Mitigation by Correcting the Distribution of Neural Activations
Figure 3 for Backdoor Mitigation by Correcting the Distribution of Neural Activations
Figure 4 for Backdoor Mitigation by Correcting the Distribution of Neural Activations
Viaarxiv icon

Improved Activation Clipping for Universal Backdoor Mitigation and Test-Time Detection

Aug 08, 2023
Hang Wang, Zhen Xiang, David J. Miller, George Kesidis

Figure 1 for Improved Activation Clipping for Universal Backdoor Mitigation and Test-Time Detection
Figure 2 for Improved Activation Clipping for Universal Backdoor Mitigation and Test-Time Detection
Figure 3 for Improved Activation Clipping for Universal Backdoor Mitigation and Test-Time Detection
Figure 4 for Improved Activation Clipping for Universal Backdoor Mitigation and Test-Time Detection
Viaarxiv icon

UMD: Unsupervised Model Detection for X2X Backdoor Attacks

Jun 02, 2023
Zhen Xiang, Zidi Xiong, Bo Li

Figure 1 for UMD: Unsupervised Model Detection for X2X Backdoor Attacks
Figure 2 for UMD: Unsupervised Model Detection for X2X Backdoor Attacks
Figure 3 for UMD: Unsupervised Model Detection for X2X Backdoor Attacks
Figure 4 for UMD: Unsupervised Model Detection for X2X Backdoor Attacks
Viaarxiv icon

Universal Post-Training Backdoor Detection

May 13, 2022
Hang Wang, Zhen Xiang, David J. Miller, George Kesidis

Figure 1 for Universal Post-Training Backdoor Detection
Figure 2 for Universal Post-Training Backdoor Detection
Figure 3 for Universal Post-Training Backdoor Detection
Figure 4 for Universal Post-Training Backdoor Detection
Viaarxiv icon

Post-Training Detection of Backdoor Attacks for Two-Class and Multi-Attack Scenarios

Jan 20, 2022
Zhen Xiang, David J. Miller, George Kesidis

Figure 1 for Post-Training Detection of Backdoor Attacks for Two-Class and Multi-Attack Scenarios
Figure 2 for Post-Training Detection of Backdoor Attacks for Two-Class and Multi-Attack Scenarios
Figure 3 for Post-Training Detection of Backdoor Attacks for Two-Class and Multi-Attack Scenarios
Figure 4 for Post-Training Detection of Backdoor Attacks for Two-Class and Multi-Attack Scenarios
Viaarxiv icon

Test-Time Detection of Backdoor Triggers for Poisoned Deep Neural Networks

Dec 06, 2021
Xi Li, Zhen Xiang, David J. Miller, George Kesidis

Figure 1 for Test-Time Detection of Backdoor Triggers for Poisoned Deep Neural Networks
Figure 2 for Test-Time Detection of Backdoor Triggers for Poisoned Deep Neural Networks
Figure 3 for Test-Time Detection of Backdoor Triggers for Poisoned Deep Neural Networks
Figure 4 for Test-Time Detection of Backdoor Triggers for Poisoned Deep Neural Networks
Viaarxiv icon

Detecting Backdoor Attacks Against Point Cloud Classifiers

Oct 20, 2021
Zhen Xiang, David J. Miller, Siheng Chen, Xi Li, George Kesidis

Figure 1 for Detecting Backdoor Attacks Against Point Cloud Classifiers
Figure 2 for Detecting Backdoor Attacks Against Point Cloud Classifiers
Figure 3 for Detecting Backdoor Attacks Against Point Cloud Classifiers
Viaarxiv icon