Trembling triggers: exploring the sensitivity of backdoors in DNN-based face recognition

• Published in: EURASIP Journal on Information Security Vol. 2020; no. 1; pp. 1 - 15
• Main Authors: Pasquini, Cecilia, Böhme, Rainer
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 23.06.2020; Springer Nature B.V; SpringerOpen
• Subjects: Adversarial machine learning; Artificial neural networks; Backdoor attacks; Communications Engineering; Dependable deep learning for security-oriented applications; Engineering; Face recognition; Facial recognition technology; Image processing; Inference; Machine learning; Networks; Neural networks; Object recognition; Security Science and Technology; Signal processing; Signal,Image and Speech Processing; Systems and Data Security; Training; Neural networks; Adversarial machine learning; Backdoor attacks

Backdoor attacks against supervised machine learning methods seek to modify the training samples in such a way that, at inference time, the presence of a specific pattern (trigger) in the input data causes misclassifications to a target class chosen by the adversary. Successful backdoor attacks have...