Decentralized Machine Learning Models to Preserve Data Privacy in Intrusion Detection Systems (IDS)

Abstract: As Technology keeps advancing rapidly, the risk of data leakage is on the increase. centralized servers are often attacked in order to control the flow of data and divert the actual network direction and breach data privacy. Data can be manipulated on a higher scale once the centralized systems have been attacked and the existing IDS fails to recognize a breach in the network server. In order to have a more robust privacy of data and accurate functioning of the IDS through ML and DL approaches to curb cyber-attacks and network attacks, the use of a decentralized ML model would be effective. This approach aims at introducing a decentralized network where data will be shared in protective nodes and each node will have an existing IDS that will not be exhausted with a huge workload. Federated Learning, which is a sub domain of DML, would offer solutions by enhancing local models on edge devices without the need and reliance of a centralized server [16]. This approach will reduce the risk of data exposure by ensuring that data stays on the source device and can only be accessed and controlled from that source only. The outcome of this research would vividly outline the effectiveness of decentralization of data and the efficiency of IDS on specialized points of the data network to protect the exchange and control of data without a maximum risk of data leak.