Date: Thursday September 16th, 2021 

Time:  2:30pm to 4:00pm 

Meeting URL:  https://us06web.zoom.us/j/83187163915?from=addon 

Passcode: If interested in attending this event, contact the Graduate Secretary days in advance of the event at csgradinfo@uwindsor.ca

 

Deep neural networks (DNN) and convolutional neural networks (CNN) have lately been achieving state-of-the-art performance on a wide range of real-life applications. However, recent work has demonstrated that deep neural networks are vulnerable to adversarial attacks, that is inputs that are almost invariant to the human eye from natural data and yet classified incorrectly by the network. Although adversarial training significantly improves model robustness, it eventually becomes a whack-a-mole game where attackers and defenders are just trying to one-up each other. Recent developments in computer applications make security aspects of machine learning increasingly important. With that in mind, an intuitive research question comes to mind, “How can we build deep neural networks that are robust to adversarial inputs?”. In this paper I introduce the first ever attempt to detect first order adversarial attacks using Siamese Neural Networks (SNN). 

 

Keywords: Adversarial Attacks, Siamese Neural Networks, GAN, Deep learning, Pairwise learning, Triplet loss 

 

Internal Reader: Dr Sherif Saad 

External Reader: Dr. Mohammad Hassanzadeh    

Advisor: Dr.Alioune Ngom