A Survey on Machine Learning Approaches for Intrusion Detection in Cloud Computing Environments for Improving Routing Payload Security and Network Privacy

Distributed denial-of-service (DDoS) attacks can significantly impair the network's availability, and recent statistics show that 30% of cyber-attacks are based on network intrusion, with 27% of incidents occurring in the financial sector. To effectively monitor large volumes of sensitive data in real-time and actively defend against these attacks, an efficient intrusion detection system (IDS) is essential. Considering the high volume of data as data-centric and Intrusion Detection (ID) as a data analysis process, Machine Learning (ML) models are capable of detecting nonconforming/ atypical network traffic through the application of signature and anomaly-based methodologies. Signature based method is prone to false negatives while anomaly-based method suffers from false positive rates. As attackers continuously evolve and create innovative malware injection techniques with new malware patterns and behaviors, there is a high demand for adaptive rule-based ML models for anomaly detection. In this paper, we attempt to rigorously review the latest trends in supervised and semi-supervised ML approaches for network traffic analysis to detect anomaly-based DDoS attacks. As the volume of the ML literature is overwhelming, our interest is specifically centered around eliciting some of the best suitable ML models that are in practice for traditional networks, IoT, Cloud, and SDN environments and can be adopted for the ensemble approach to construct adaptive rule-based ML algorithms. We believe such a consolidated overview for a broad set of platforms will help readers to quickly make a choice on the required platforms, datasets involved, and methods that suit their applications.