A Study on Machine Learning and Cryptographic based Intrusion Detection System in Blockchain and Data Science Environment

Abstract

The recent speedy development of Internet of Things, vehicular networks and cloud computing has greatly boosted the need of effective and safe intrusion detection systems. The shortcomings of the traditional IDS-based approaches are centralized architecture, issue of scalability, and inability to analyze the encrypted and distributed data. To address these issues, the recent studies have paid attention to the incorporation of the latest technologies such as machine learning, deep learning, cryptographic solutions, federated learning, and blockchain. In this paper, a review of the IDS frameworks used in modern times, especially focusing on cryptographic IDS and blockchain-based federated learning IDS. The cryptographic techniques will ensure the confidentiality of the information, integrity, and authentication of the data, and ML and DL models will raise awareness of more advanced and evolving cyber threats. Federated learning can be used to train models in a decentralized fashion, using no raw data, preserving privacy, and blockchain provides transparency, immutability, and trust in a distributed environment. This paper compares the existing approaches in regards to methodology, performance measures and applications and places and describes the merits and demerits of the approaches. The key issues such as computational overheads, the latency of communication, scalability issues, and the vulnerability to the poisoning attacks are also discussed, and potential future research opportunities are also outlined to develop efficient and secure solutions to the IDS.