A Blockchain-assisted Collaborative Ensemble Learning for Network Intrusion Detection

Abstract

With the rapid growth of the Internet network, cyber attacks (mainly DDOS, U2R, Probe, Infiltration and Heart-bleed etc) on networks and computer systems have also increased expeditiously. Intrusion detection system has proven to be one of the most effective methods to resist cyber attacks. Most traditional machine learning methods can only learn shallow features of the data, so they are weak in detecting complex data. Recently, ensemble learning has begun to be applied in network intrusion detection due to their excellent generalization ability. However, most of them are designed based on machine learning methods. They usually do not have multiple hidden layers, which will lead to low detection accuracy. Furthermore, when the base learners in ensemble learning make voting decisions, the data can be easily tampered, which incurs wrong detection results. To solve the above problems, we propose an intrusion detection method based on blockchain and collaborative ensemble learning. In detail, to address the low detection rate of machine learning model, we design an ensemble deep learning approach combining with a weighted dynamic voting mechanism, which can enhance base learners with excellent performance and weaken base learners with poor performance. To solve the problem of data tampering during individual model voting decisions, we explore blockchain to verify the detection results of the individual model and the final results of the voting. Finally, we evaluate the performance of the proposed system on the CICIDS-2017 dataset. The experimental results demonstrate the accuracy and effectiveness of our proposed system.