Adversarial Attacks and Defences of Various Artificial Intelligent Models
Tanmay Singh, Parijat Rai, Saumil Sood, Siddhant Nigam, Suchi Kumari · 2023
Adversarial attacks are becoming increasingly common and sophisticated as time passes and new defenses are being researched and developed. Adversarial machine learning is one of the techniques used to attack artificial intelligence models by feeding deceptive data. Therefore, in this paper, adversarial attacks against machine learning and deep learning are studied, and some defense strategies are also proposed against them. Two attack strategies; Evasion and Poison are used to attack the Support Vector Machine (SVM) model and Convolution Neural Network (CNN) model, respectively. Some defense mechanisms are applied to safeguard the models from such attacks. The models are applied and tested on the MNIST dataset. The results obtained from the models, before and after the defenses are applied are compared.