Multi-Class Triplet Loss With Gaussian Noise for Adversarial Robustness
| Autor: | Muhammed Amin Abdullah, Zhiguang Qin, Edward Yellakuor Baagyere, Benjamin Appiah, Kwabena Owusu-Agyemang |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
General Computer Science
Computer science Iterative method metric learning 010501 environmental sciences 01 natural sciences Data modeling Adversarial detection symbols.namesake Adversarial system Triplet loss Robustness (computer science) 0502 economics and business General Materials Science 050207 economics Computer Science::Cryptography and Security 0105 earth and related environmental sciences Artificial neural network business.industry 05 social sciences General Engineering Pattern recognition anomaly detection Gaussian noise adversarial training symbols lcsh:Electrical engineering. Electronics. Nuclear engineering Artificial intelligence business lcsh:TK1-9971 Classifier (UML) |
| Zdroj: | IEEE Access, Vol 8, Pp 171664-171671 (2020) |
| ISSN: | 2169-3536 |
| Popis: | Deep Neural Networks (DNNs) classifiers performance degrades under adversarial attacks, such attacks are indistinguishably perturbed relative to the original data. Providing robustness to adversarial attacks is an important challenge in DNN training, which has led to extensive research. In this paper, we harden DNN classifiers under the adversarial attacks by regularizing their deep internal representation space with Multi-class Triplet regularization method. This method enables DNN classifier to learn a feature representation that detects similarities between adversarial and clean images and brings similar images close to their original class and pushes dissimilar images away from their false classes. This training process with our Multi-class Triplet regularization method in combination with Gaussian noise injection proves to be more robust in detecting adversarial attacks exceeding that of adversarial training on strong iterative attacks. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|