With the rising complexity of modern information systems and the resulting ever increasing flow of big data, the benefits of Artificial Intelligence (AI) are now widely recognized.
Specifically, Machine Learning (ML) methods are already deployed to solve diverse real-world tasks – especially with the advent of deep learning. Fascinating examples of practical achievements of ML are machine translation, travel and vacation recommendations, object detection and tracking, and even various applications in healthcare. Furthermore, ML is rightly considered to be a technology enabler, as it has shown great potential in the context of telecommunication systems or autonomous driving.
Nevertheless, modern society is increasingly relying on Information Technology (IT) systems – including autonomous ones – which are also actively leveraged by malicious entities.
Digital threats are, in fact, continuously evolving, and some researchers believe attackers will have sufficient capabilities to harm or kill humans by 2025. To prevent such incidents and mitigate the plethora of risks that can target current and future IT systems, defensive mechanisms require the capability to quickly adapt to the (i) mutating environments and (ii) dynamic threat landscape. Coping with such a twofold requirement via static and human-defined methods is clearly unfeasible, and deployment of Machine Learning in cybersecurity is inescapable.
(https://dl.acm.org. Adaptad)
