Practical Challenges of Applying Machine Learning in Cybersecurity

Presented by Sherif Saad as a part of the 2020 Serene-risc Workshop on The State of Canadian Cybersecurity Conference: Human-Centric Cybersecurity.

About the presentation

Currently, Artificial Intelligence (AI) is transforming the world’s industries. It is expected that the next generation of cybersecurity products will incorporate Artificial Intelligence as a core security mechanism. However, compared to other sectors, the cybersecurity industry faces unique challenges for AI practitioners and pioneers. Cybersecurity vendors are adding different AI technologies to their security products. However, the massive number of cyber attacks indicates that AI is still in the early stages of practical use in cybersecurity.  For those reasons, applying AI technologies in cybersecurity in the wild requires unique tuning and engineering to utilize the full potential of AI.  In this presentation, we discuss practical challenges that AI practitioners need to address when applying machine learning in cybersecurity. We demonstrate these challenges using a walk-through example on engineering a behavioural-based biometric access control system using different machine learning techniques.  We show how using machine-learning best practice techniques is not always possible or suitable for the cybersecurity production environment. The experimental results in production environments show a significant gap between production and lab settings results. Finally, we show that these challenges can be applied to other cybersecurity areas such as malware detection and discuss some possible solutions and future work.

About the speaker

Sherif is an assistant professor of cybersecurity at the University of Windsor and found of the WASP (Windsor Advanced Security and Privacy) research lab that develops innovative and usable security solutions for unconventional cybersecurity threats. His research interests include cybersecurity, applied machine learning, and software engineering. With WASP labs, Dr.Saad is leading several research projects with NRC, Canada DND, NSERC, MITACS, and other private organizations. Dr.Saad holds several research grants with totalling over $700,000 (CAD).  He has published several articles in prestigious and top-tier computing and cybersecurity journals and conferences. Dr. Saad has 10+ years of industry experience in cybersecurity and applied machine learning. During these years, he had the following roles: software developer, application security engineer, software security architect, chief software architect, and director of engineering. He worked with many companies to develop security systems for clients in the defence and finance sectors. Some of those clients include MasterCard, American Express, US DoD, Booz Allen, RCMP, and DRDC.