Affiliations: Graduate School of Engineering, The University of Tokyo, 7 Chome-3-1 Hongo, Bunkyo, Tokyo, Japan. E-mails: [email protected], [email protected]
Abstract: The most common methods used in cyber attack detection are signature scan and anomaly detection. In the case of applying these approaches, a countermeasure against an upcoming cyber attack is made only if a signature of cyber attack or an anomaly is detected. That means cyber defense systems encounter cyber attacks with no preparation, and our study focuses on this problem. This time, we attempt to discover the useful social data for the prediction of cyber attack motivation and opportunity. For the prediction of cyber attack motivation, the news articles were used as the dataset. As a result, using Artificial Neural Networks and the core keywords extracted from the news articles directly correlated to a cyber attack or the news articles not correlated to cyber attack brought better precision/recall. For the prediction of cyber attack opportunity, the security vulnerability feeds were used as the dataset. The precision/recall of the prediction result was better when using the core keywords as the feature and Artificial Neural Networks as the prediction algorithm.
Keywords: Cyber attack prediction, social data analysis, natural language processing, news article analysis, Twitter analysis, security vulnerability feeds analysis, SVM classification, artificial neural networks, convolutional neural networks
