Privacy-Preserving Collaborative Intrusion Detection Systems: A Federated Learning Framework
Chukwuebuka Ezelu, Ulrich Buehler · 2022
Security concerns are one of the significant obstacles to collaborative systems. With the ever-increasing sophistication of attacks in networked environments, it is impossible to overlook the need for collaboration. Collaborative Intrusion Detection Systems (CIDS) presents a solution to combating such attacks but also introduces some privacy concerns. The need for information transmission within CIDS architecture raises some privacy issues, which can be deterring to organizations with sensitive data. This work presents a forest-based Federated Learning framework that allows organizations to collaboratively build a forest without compromising the security and privacy of their data. Two strategies were proposed for building the forest and were evaluated against a centralized dataset. The result shows that the proposed framework outperforms the centralized dataset framework based on model reliability and accuracy of prediction.