Abstract
The rapid adoption of Open Radio Access Network (Open-RAN) architectures has brought unprecedented innovation opportunities in modern telecommunications networks. However, this evolution also introduces novel security challenges, particularly in demanding scenarios where swift decision-making is critical. In this paper, we conduct an in-depth investigation into model poisoning attacks in ensemble learning, highlighting their implications for network security, and provide a detailed demonstration of our proposed Open-RAN Intrusion Detection System (IDS), which is seamlessly incorporated into the security module of the near Real-Time RAN Intelligent Controller (nearRT-RIC). The strategic placement of the IDS within the nearRT-RIC ensures its operation within the demanding 10 ms to 1 second control loop range, enabling nearRT intrusion detection capabilities. Through rigorous evaluation and experimentation, our solution showcases promising results in enhancing network security without compromising performance.