Resumen de Deep Learning Applications on Cybersecurity
Carlos Lago, Rafael Romón, Iker Pastor López, Borja Sanz Urquijo, Alberto Tellaeche Iglesias, Pablo García Bringas
Security has always been one of the biggest challenges faced by computer systems, recent developments in the field of Machine Learning are affecting almost all aspects of computer science and Cybersecurity is no different. In this paper, we have focused on studying the possible application of deep learning techniques to three different problems faced by Cybersecurity: SPAM filtering, malware detection and adult content detection in order to showcase the benefits of applying them. We have tested a wide variety of techniques, we have applied LSTMs for spam filtering, then, we have used DNNs for malware detection and finally, CNNs in combination with Transfer Learning for adult content detection, as well as applying image augmentation techniques to improve our dataset. We have managed to reach an AUC over 0.9 on all cases, demonstrating that it is possible to build cost-effective solutions with excellent performance without complex architectures.
