Incident Detection over Unified Threat Management Platform on a Cloud Network

Abstract

Artificial Intelligence (AI) techniques provide many intelligent methods for security solutions in various domains such as finance, networking, cloud computing, health records and individual's identity. AI achieves security mechanisms like antivirus, firewalls, intrusion detection system (IDS) and cryptography by using machine learning methods and data analysis techniques. As the modern AI techniques help improving security systems, criminal activities are also becoming updated simultaneously. Machine learning methods along with data analysis tools have become popular to prevent security systems from threats and hacking activities. This work contributes to secure cloud networks and help them prevent malicious attacks. In this paper, Bidirectional long short-term memory (BLSTM) is used to detect incidents over unified threat management (UTM) platform operated on cloud network. Results are compared with K-nearest neighbor which is a baseline technique. Time series input samples recorded over UTM platform are used for training and testing purposes. We obtain accuracy score of 98.47% with 0.0186 mean squared error (MSE) using KNN while BLSTM provides 98.6% accuracy score with 0.002 loss, which is better than the KNN.