Linkbreaker: Breaking the backdoor-trigger link in DNNs via neurons consistency check

• Main Authors: CHEN, Zhenzhu, WANG, Shang, FU, Anmin, GAO, Yansong, YU, Shui, DENG, Robert H.
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2022
• Subjects: Neurons; Training; Predictive models; Data models; Feature extraction; Artificial intelligence; Training data; Backdoor attack; defense; deep learning; AI security; Information Security
• Online Access: https://ink.library.smu.edu.sg/sis_research/7250
• Institution: Singapore Management University

Backdoor attacks cause model misbehaving by first implanting backdoors in deep neural networks (DNNs) during training and then activating the backdoor via samples with triggers during inference. The compromised models could pose serious security risks to artificial intelligence systems, such as misi...