AI-based systems are already widespread and highly influential for many aspects of our life. Only some of the domains already fully reshaped with AI are: automotive industry, retail, e-commerce, banking and financial services. In the last few years, we have also witnessed how AI has found many applications within the computer security domain, and it is one of the driving technologies behind malware detection and classification, phishing detection, and is enabling using biometrics for authentication purposes. Some of the main reasons we use AI in the security domain are its ability to analyze huge amounts of data and to automatically discover complex patterns which would otherwise go undetected.
While AI remains an excellent tool for coping with many security problems, with more widespread adoption of it, we should also be more aware of potential problems, intrinsic properties of AI systems, and in general good and bad practices when working with them.
In this talk, these common problems with machine learning for security will be shared, together with specifics of machine learning methods for security and recommendations on best practices.