Backdoor Threat

Backdoor attacks exploit vulnerabilities in machine learning models, causing them to produce specific, malicious outputs when presented with a hidden "trigger," while appearing normal otherwise. Current research focuses on understanding and mitigating these attacks across various model architectures, including those used in federated learning and natural language processing, with a particular emphasis on identifying effective defenses and developing robust evaluation metrics. The significance of this research lies in ensuring the trustworthiness and security of machine learning systems deployed in critical applications, where compromised models could have severe consequences.

Papers

June 10, 2024

Lurking in the shadows: Unveiling Stealthy Backdoor Attacks against Personalized Federated Learning
Xiaoting Lyu, Yufei Han, Wei Wang, Jingkai Liu, Yongsheng Zhu, Guangquan Xu, Jiqiang Liu, Xiangliang Zhang
Backdoor Attack Personalized Federated Learning Quantum Shadow Stealthy Backdoor Attack Effective Backdoor Attack Hidden Backdoor Backdoor Threat

April 17, 2024

The Victim and The Beneficiary: Exploiting a Poisoned Model to Train a Clean Model on Poisoned Data
Zixuan Zhu, Rui Wang, Cong Zou, Lihua Jing
Backdoor Attack Poisoning Attack Poisoned Data Feature Suppression Male Victim Backdoor Threat

February 21, 2024

Backdoor Attacks on Dense Passage Retrievers for Disseminating Misinformation
Quanyu Long, Yue Deng, LeiLei Gan, Wenya Wang, Sinno Jialin Pan
Backdoor Attack Retrieval Augmented Language Model Dense Retriever Dense Passage Retrieval Online Misinformation Misinformation Campaign Backdoor Threat

November 25, 2023

Effective Backdoor Mitigation Depends on the Pre-training Objective
Sahil Verma, Gantavya Bhatt, Avi Schwarzschild, Soumye Singhal, Arnav Mohanty Das, Chirag Shah, John P Dickerson, Jeff Bilmes
Backdoor Attack Backdoor Defense Backdoor Removal Backdoor Behavior Backdoor Threat

August 31, 2023

Everyone Can Attack: Repurpose Lossy Compression as a Natural Backdoor Attack
Sze Jue Yang, Quang Nguyen, Chee Seng Chan, Khoa D. Doan
Backdoor Attack Backdoor Defense Lossy Compression Lossy Image Compression Stealthy Backdoor Attack Backdoor Threat Natural Backdoor

May 16, 2023

UOR: Universal Backdoor Attacks on Pre-trained Language Models
Wei Du, Peixuan Li, Boqun Li, Haodong Zhao, Gongshen Liu
Pre Trained Language Model Backdoor Attack Universal Backdoor Backdoor Threat

February 9, 2023

Mithridates: Auditing and Boosting Backdoor Resistance of Machine Learning Pipelines
Eugene Bagdasaryan, Vitaly Shmatikov
Backdoor Attack Backdoor Defense Machine Learning Pipeline Audit Evidence Backdoor Threat