Nowadays, Deep neural networks are the most popular approach which we see its usage in different applications and tasks. As day growth its usage in different tasks, checking the vulnerability of these networks is being a very important fundamental issue. Therefore, analyzing of each machine learning model (such as neural network) for its vulnerability, is a useful task to assess the usage of that in critical situations. In this session, We try to cover the key definition step’s of vulnerability of deep neural networks and its defense strategies against simplest vulnerability at first. Then when the minds are boiled, we try to implement and test them in a practical manner. Also, covering a teamwork remote session for more collaboration is available at the end of the session.
Ph.D candidate at Amirkabir University of Technology
Mohammad Khalooei is a Ph.D candidate at Amirkabir University of Technology (Tehran Polytechnic) in the department of computer engineering. He works at the Laboratory of Intelligence and Multimedia Processing of AUT. He is interested in artificial intelligence fields and working on vulnerability of deep neural network, adversarial machine learning and unsupervised learning in theoretical and also deployment phases. He also has some experiences in counseling on using deep neural networks for real data processing and joint international project.