Deep learning for predictive alerting and cyber-attack mitigation

The successful security management of ICT systems and services is essential for an effective cyber security posture. The main objective is to minimize and control the damage caused by cyber-attacks and incidents, to provide effective response and recovery, and to invest efforts in preventing future cyber incidents. To achieve this objective, cyber threat intelligence (CTI) is widely applied, as it is considered a crucial mechanism to proactively defend against modern and dynamically evolving cyber threats and attacks. However, there are multiple challenges in the field of CTI, as there is an enormous amount of unstructured threats data in cyberspace that needs to be collected, classified, analyzed, and shared between states, organizations, or companies. Facing this challenge, data mining techniques and machine learning algorithms are essential for providing meaningful CTI information due to their ability to extract indistinct and hidden patterns in the data. Based on data mining techniques and machine learning algorithms' potential for successfully implementing cyber threat intelligence services, this paper develops an efficient predictive alerting model in a threat intelligence engine using the Deep Residual Network (DRN) model. Further, the main goal is to compare the performance of the DRN model with other machine learning models such as Sequential Rule Mining, IntruDTree, ScaleNet, etc. According to our experimental results, the DRN outperformed other tested machine learning models by achieving better results on parameters such as precision, recall, and F-measure.