Deep Neural Networks for Cybersecurity

Le résumé fourni par la source

Traditional cyber security approaches involving signature-based threat detection and dynamic analysis in a sandboxed environment can only identify known threats once the originating source is identified and explored. Even after that is done, a sandbox can only store a limited number of signatures, and hackers have already found techniques to evade sandboxes. A neural network’s ability to generalize and infer applicative representations without imposing any restrictions or fixed relationships on the input data makes it an extremely strong contender for modeling and detecting a variety of threats, even some seemingly novel ones. Proposed applications include intrusion detection and prevention systems, network traffic and user behavior analysis, along with spam, social engineering and malware detection. This chapter explores recent advancements made by deep learning applications in the cyber security landscape and its applicability in addressing challenges. The chapter ends by detailing future prospects in the domain.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.