Malware is one of the most relevant problems in cybersecurity. The Internet spreads tons of malicious software, compromising several personal devices. Big data models, based on machine learning, can handle these big quantities of malicious information, but machine learning algorithms were not designed to deal with adversaries, and this issue is generating a big gap between confidence and efficiency that has not been filled yet. This talk aims to introduce vulnerabilities on current machine learning based solutions, and different scenarios where adversaries exploit them. It also aims to give some advice for strengthening machine learning models against adversaries, with the aim of helping to solve this open problem.

Héctor Menéndez is Research Associate in University College London. He holds a PhD. in Computer Science in Universidad Autónoma de Madrid (2014), a MSc in Computer Science (2012), a MSc in Mathematics (2013), a BSc in Computer Science (2010) and a BSc in Mathematics from Universidad Autónoma de Madrid (2010). He is involved with Software System Engineering Group (SSE) at the Department of Computer Science in UCL. His main research interests are related to Information Theory, Software Testing, Malware Analysis, Adversarial Machine Learning, Clustering, Graph-based algorithms and Evolutionary Algorithms.