Addressing Poisoning Attacks Against Federated Intrusion Detection Systems in Internet of Healthcare Things

Federated learning emerges as a key technology, allowing decentralized model training in Internet of Healthcare Things (IoHT). This aims to preserve data privacy and accommodates the dynamic nature of health data, particularly in the development of sensitive systems such as Intrusion Detection Systems (IDS), known as Federated Intrusion Detection Systems (FIDS). Nevertheless, this approach is susceptible to adversarial poisoning attacks, which involve the insidious introduction of tainted data into federated models. This paper highlights the challenge of poisoning attacks against FIDS in the IoHT ecosystem. We introduce a threat scenario illustrating the potential vulnerabilities and implications of such attacks. The scenario emphasizes the need to address this nuanced cybersecurity challenge in the ongoing development and deployment of IoHT technologies. We evaluate our proposals through rigorous experiments, examining their effectiveness in simulated attack scenarios.