Neural Backdoor

Neural backdoors are malicious modifications to deep neural networks (DNNs) that cause them to misclassify inputs containing a specific trigger, compromising model integrity and security. Current research focuses on developing robust defense mechanisms, primarily employing neuron pruning techniques guided by weight changes, neuron magnitude, or optimization algorithms like reinforcement learning, often combined with unlearning and relearning strategies. These efforts aim to identify and remove backdoored neurons or mitigate their effects without significantly impacting the model's overall accuracy, addressing a critical vulnerability in the deployment of DNNs across various applications. The significance lies in ensuring the trustworthiness and reliability of AI systems, particularly in security-sensitive contexts.

Papers

September 29, 2024

Psychometrics for Hypnopaedia-Aware Machinery via Chaotic Projection of Artificial Mental Imagery
Ching-Chun Chang, Kai Gao, Shuying Xu, Anastasia Kordoni, Christopher Leckie, Isao Echizen
Backdoor Attack Backdoor Trigger Random Projection Psychometric Property Mental Imagery Neural Backdoor

May 30, 2024

Unveiling and Mitigating Backdoor Vulnerabilities based on Unlearning Weight Changes and Backdoor Activeness
Weilin Lin, Li Liu, Shaokui Wei, Jianze Li, Hui Xiong
Backdoor Defense Weight Freezing Neural Backdoor Backdoor Effect Resilient Backdoor

May 28, 2024

May 23, 2024

Unified Neural Backdoor Removal with Only Few Clean Samples through Unlearning and Relearning
Nay Myat Min, Long H. Pham, Jun Sun
Backdoor Attack Harmful Unlearning Backdoor Removal Neural Backdoor

April 22, 2024

Dual Model Replacement:invisible Multi-target Backdoor Attack based on Federal Learning
Rong Wang, Guichen Zhou, Mingjun Gao, Yunpeng Xiao
Backdoor Attack Neural Backdoor Invisible Backdoor Attack Multimodal Backdoor

February 10, 2024

Architectural Neural Backdoors from First Principles
Harry Langford, Ilia Shumailov, Yiren Zhao, Robert Mullins, Nicolas Papernot
Backdoor Attack First Principle Neural Backdoor Backdoor Behavior Resilient Backdoor

May 27, 2023

Backdooring Neural Code Search
Weisong Sun, Yuchen Chen, Guanhong Tao, Chunrong Fang, Xiangyu Zhang, Quanjun Zhang, Bin Luo
Vulnerable Code Code Search Code Vulnerability Neural Backdoor Malicious Code

May 24, 2023

Reconstructive Neuron Pruning for Backdoor Defense
Yige Li, Xixiang Lyu, Xingjun Ma, Nodens Koren, Lingjuan Lyu, Bo Li, Yu-Gang Jiang
Backdoor Defense Backdoor Detection Neuron Pruning Backdoor Removal Neural Backdoor

April 24, 2023

Enhancing Fine-Tuning Based Backdoor Defense with Sharpness-Aware Minimization
Mingli Zhu, Shaokui Wei, Li Shen, Yanbo Fan, Baoyuan Wu
Backdoor Attack Sharpness Aware Minimization Backdoor Defense Neural Backdoor Backdoor Effect

April 12, 2023

Exploiting Logic Locking for a Neural Trojan Attack on Machine Learning Accelerators
Hongye Xu, Dongfang Liu, Cory Merkel, Michael Zuzak
Logic Reasoning Phantom 2D Accelerator Machine Learning Accelerator Neural Backdoor Neural Trojan Trigger Attack

October 21, 2022

Neural Architectural Backdoors
Ren Pang, Changjiang Li, Zhaohan Xi, Shouling Ji, Ting Wang
Neural Architecture Search Attack Vector Hidden Backdoor Neural Backdoor Trigger Design

July 10, 2022

One-shot Neural Backdoor Erasing via Adversarial Weight Masking
Shuwen Chai, Jinghui Chen
Deep Neural Network Training Data Adversarial Mask Backdoor Removal Neural Backdoor

April 20, 2022

Backdooring Explainable Machine Learning
Maximilian Noppel, Lukas Peter, Christian Wressnegger
Explainable Machine Learning Malware Classification Backdoor Learning Stealthy Attack Neural Backdoor

February 13, 2022

Training with More Confidence: Mitigating Injected and Natural Backdoors During Training
Zhenting Wang, Hailun Ding, Juan Zhai, Shiqing Ma
Training Data Backdoor Learning Neural Backdoor Natural Backdoor