Combining deep learning with rough set analysis: A model of cyberspace situational awareness

Abstract

A cyberspace situational awareness (CSA) model, DLRSA model, is proposed in this paper, with feature extraction based on deep learning (DL) and situation assessment based on rough set analysis (SARSA). We focus on network flow instead of server logs of IDS to extract features, transforming source data into information and establishing knowledge base. On account of Gaussian-Bernoulli deep belief network (GBDBN), accurate feature information is provided for assessment. While DLRSA model assesses situation in reference to pattern recognition, assessment strategy could be given by rough set analysis and pattern information abstracted deeply. Experiments indicate that DLRSA model has low extraction error and succinct assessment rule.