An Efficient Network IDS for Cloud Environments Based on a Combination of Deep Learning and an Optimized Self-adaptive Heuristic Search Algorithm

Nowadays, Cloud Computing (CC) is one of the fastest emerging core technologies in the current information era. It is leading a new revolution on the ways of data storage and calculation. CC remains gaining traction among organizations thanks to its appealing features like pay-per-use model for billing customers, elasticity, ubiquity, scalability and availability of resources for businesses. Hence, many organizations are moving their workloads or processes to cloud due to its inherent advantages. Nevertheless, several security issues arise with the transition to this computing paradigm including intrusion detection. Attackers and intruders developed new sophisticated tools defeating traditional Intrusion Detection Systems (IDS) by huge amount of network traffic data and dynamic behaviors. The existing Cloud IDSs suffer from low detection accuracy and high false positive rate. To overcome this issue, we propose a smart approach using a self-adaptive heuristic search algorithm called “Improved Self-Adaptive Genetic Algorithm” (ISAGA) to build automatically a Deep Neural Network (DNN) based Anomaly Network Intrusion Detection System (ANIDS). ISAGA is a variant of standard Genetic Algorithm (GA), which is developed based on GA improved through an Adaptive Mutation Algorithm (AMA) and optimization strategies. The optimization strategies carried out are Parallel Processing and Fitness Value Hashing that reduce execution time, convergence time and save processing power. Our approach consists of using ISAGA with the goal of searching the optimal or near optimal combination of most relevant values of the parameters included in construction of DNN based IDS or impacting its performance, like feature selection, data normalization, architecture of DNN, activation function, learning rate and Momentum term, which ensure high detection rate, high accuracy and low false alarm rate. CloudSim 4.0 simulator platform and CICIDS2017 dataset were used for simulation and validation of the proposed system. The implementation results obtained have demonstrated the ability of our ANIDS to detect intrusions with high detection accuracy and low false alarm rate, and have indicated its superiority in comparison with state-of-the-art methods.