Backdoor Defense

Backdoor attacks subtly inject malicious functionality into machine learning models during training, causing them to misbehave when presented with specific "triggers." Current research focuses on developing robust defenses against these attacks across various model architectures, including graph neural networks, large language models, diffusion models, and multimodal contrastive learning models like CLIP, employing techniques such as trigger inversion, model unlearning, and data augmentation. Effective backdoor defense is crucial for ensuring the reliability and security of machine learning systems in diverse applications, ranging from autonomous driving to cybersecurity, and is a significant area of ongoing investigation within the broader field of adversarial machine learning.

Papers

January 11, 2025

SafeSplit: A Novel Defense Against Client-Side Backdoor Attacks in Split Learning
Phillip Rieger, Alessandro Pegoraro, Kavita Kumari, Tigist Abera, Jonathan Knauer, Ahmad-Reza Sadeghi
LeArning Abstract Backdoor Attack Split Learning Backdoor Defense Split and Fit Defense Method

January 10, 2025

December 29, 2024

Cut the Deadwood Out: Post-Training Model Purification with Selective Module Substitution
Yao Tong, Weijun Li, Xuanli He, Haolan Zhan, Qiongkai Xu
Backdoor Attack Backdoor Defense Dynamic ModulE Selective Engagement Diffusion Purification Image Purification

December 3, 2024

Gracefully Filtering Backdoor Samples for Generative Large Language Models without Retraining
Zongru Wu, Pengzhou Cheng, Lingyong Fang, Zhuosheng Zhang, Gongshen Liu
Full Model Backdoor Attack Backdoor Defense Generative LLM Retraining Method Backdoor Sample

December 2, 2024

CopyrightShield: Spatial Similarity Guided Backdoor Defense against Copyright Infringement in Diffusion Models
Zhixiang Guo, Siyuan Liang, Aishan Liu, Dacheng Tao
Diffusion Model Backdoor Defense Copyright Protection Defense Method Copyright Violation

November 27, 2024

Neutralizing Backdoors through Information Conflicts for Large Language Models
Chen Chen, Yuchen Sun, Xueluan Gong, Jiaxin Gao, Kwok-Yan Lam
Full Model Backdoor Attack Large Language Backdoor Defense Knowledge Conflict

November 18, 2024

CROW: Eliminating Backdoors from Large Language Models via Internal Consistency Regularization
Nay Myat Min, Long H. Pham, Yige Li, Jun Sun
Backdoor Attack Adversarial Perturbation Consistency Regularization Backdoor Defense

November 17, 2024

BackdoorMBTI: A Backdoor Learning Multimodal Benchmark Tool Kit for Backdoor Defense Evaluation
Haiyang Yu, Tian Xie, Jiaping Gui, Pengyang Wang, Ping Yi, Yue Wu
Backdoor Attack Backdoor Defense Backdoor Learning Mitigating Backdoor Backdoor Model Multimodal Benchmark Datasets

October 25, 2024

Expose Before You Defend: Unifying and Enhancing Backdoor Defenses via Exposed Models
Yige Li, Hanxun Huang, Jiaming Zhang, Xingjun Ma, Yu-Gang Jiang
Full Model Backdoor Attack Backdoor Defense Anti Unification Backdoor Detection

October 15, 2024

BeniFul: Backdoor Defense via Middle Feature Analysis for Deep Neural Networks
Xinfu Li, Junying Zhang, Xindi Ma
Network Programming Backdoor Defense Deeper Network Backdoor Removal Input Level Backdoor Detection Intermediate Feature

October 13, 2024

Uncovering, Explaining, and Mitigating the Superficial Safety of Backdoor Defense
Rui Min, Zeyu Qin, Nevin L. Zhang, Li Shen, Minhao Cheng
Line by Line Explanation Backdoor Defense Diffusion Purification Backdoor Purification

October 2, 2024

"No Matter What You Do": Purifying GNN Models via Backdoor Unlearning
Jiale Zhang, Chengcheng Zhu, Bosen Rao, Hao Sui, Xiaobing Sun, Bing Chen, Chunyi Zhou, Shouling Ji
Graph Neural Network Backdoor Attack Backdoor Defense Backdoor Learning Backdoor Removal

September 24, 2024

Adversarial Backdoor Defense in CLIP
Junhao Kuang, Siyuan Liang, Jiawei Liang, Kuanrong Liu, Xiaochun Cao
Adversarial Example Single CLIP Backdoor Defense Backdoor Sample Adversarial Backdoor Multimodal Backdoor

September 21, 2024

Obliviate: Neutralizing Task-agnostic Backdoors within the Parameter-efficient Fine-tuning Paradigm
Jaehan Kim, Minkyoo Song, Seung Ho Na, Seungwon Shin
Parameter Efficient Fine Tuning Backdoor Defense Adaptive Attack Task Agnostic Backdoor

September 20, 2024

PureDiffusion: Using Backdoor to Counter Backdoor in Generative Diffusion Models
Vu Tuan Truong, Long Bao Le
Backdoor Attack Generative Diffusion Model Backdoor Defense Backdoor Trigger

September 4, 2024

Backdoor defense, learnability and obfuscation
Paul Christiano, Jacob Hilton, Victor Lecomte, Mark Xu
Efficient Learning Backdoor Defense Obfuscation Technique PAC Learnability

September 2, 2024

Backdoor Defense through Self-Supervised and Generative Learning
Ivan Sabolić, Ivan Grubišić, Siniša Šegvić
Generative Model Self Supervised Backdoor Attack Backdoor Defense Inference Attack Generative Learning

September 1, 2024

Fisher Information guided Purification against Backdoor Attacks
Nazmul Karim, Abdullah Al Arafat, Adnan Siraj Rakin, Zhishan Guo, Nazanin Rahnavard
Backdoor Attack Backdoor Defense Fisher Information Backdoor Removal Backdoor Policy Backdoor Purification

August 28, 2024

VFLIP: A Backdoor Defense for Vertical Federated Learning via Identification and Purification
Yungi Cho, Woorim Han, Miseon Yu, Younghan Lee, Ho Bae, Yunheung Paek
Backdoor Attack Person Identification Vertical Federated Learning Backdoor Defense VFL Algorithm Horizontal Federated Learning

Backdoor Defense

Papers

SafeSplit: A Novel Defense Against Client-Side Backdoor Attacks in Split Learning

Towards Backdoor Stealthiness in Model Parameter Space

Fine-tuning is Not Fine: Mitigating Backdoor Attacks in GNNs with Limited Clean Data

Cut the Deadwood Out: Post-Training Model Purification with Selective Module Substitution

Gracefully Filtering Backdoor Samples for Generative Large Language Models without Retraining

CopyrightShield: Spatial Similarity Guided Backdoor Defense against Copyright Infringement in Diffusion Models

Neutralizing Backdoors through Information Conflicts for Large Language Models

CROW: Eliminating Backdoors from Large Language Models via Internal Consistency Regularization

BackdoorMBTI: A Backdoor Learning Multimodal Benchmark Tool Kit for Backdoor Defense Evaluation

Expose Before You Defend: Unifying and Enhancing Backdoor Defenses via Exposed Models

BeniFul: Backdoor Defense via Middle Feature Analysis for Deep Neural Networks

Uncovering, Explaining, and Mitigating the Superficial Safety of Backdoor Defense

"No Matter What You Do": Purifying GNN Models via Backdoor Unlearning

Adversarial Backdoor Defense in CLIP

Obliviate: Neutralizing Task-agnostic Backdoors within the Parameter-efficient Fine-tuning Paradigm

PureDiffusion: Using Backdoor to Counter Backdoor in Generative Diffusion Models

Backdoor defense, learnability and obfuscation

Backdoor Defense through Self-Supervised and Generative Learning

Fisher Information guided Purification against Backdoor Attacks

VFLIP: A Backdoor Defense for Vertical Federated Learning via Identification and Purification