A Novel Semi-Supervised Learning Approach for Network Intrusion Detection on Cloud-Based Robotic System

Although the cloud-based robotic system has provided the services in various industries, its data safety is continuously threatened, and the network intrusion detection system (NIDS) is considered as a necessary component to ensure its security. In recent years, many machine learning (ML) techniques have been applied for building a more intelligent NIDS. Most NIDSs based on the ML method and artificial intelligence techniques are either supervised or unsupervised. However, the supervised learning for NIDS depends much on the labeled data. This weakness makes it harder to detect the latest attack patterns. Meanwhile, the unsupervised learning for NIDS often fails to give the satisfactory results. Therefore, this paper proposed a novel fuzziness-based semi-supervised learning approach via ensemble learning for network intrusion detection on the cloud-based robotic system, which can address the above issues. First, due to the good generalization ability of ensemble learning, we construct an ensemble system trained by the labeled data. Moreover, for better utilizing the unlabeled data, a fuzziness-based method is adopted for data analysis. In this way, the noisy and redundant examples in the data set are removed. Finally, we use the same ensemble approach to combine both supervised and unsupervised parts. To verify the effectiveness and robustness of the NIDS, the proposed approach is tested on the NSL-KDD data set, which is a commonly used traffic data set. The experimental results show that the proposed approach achieves the accuracy 84.54% and 71.29% on the, respectively, “KDDTest+” and “KDDTest-21” data sets. When compared with the state-of-the-art method, the proposed method also delivers a promising result.