Evaluating Robustness and Reliability of a C-NIDS for IoT Networks in Virtualized Environments

Abstract

Protecting thousands of smart devices across multiple IoT domains requires robust and collaborative security solutions, such as Collaborative Network Intrusion Detection Systems (C-NIDS). In this context, collaboration among NIDS improves intrusion detection by sharing information about local threats. However, most current solutions neglect the reliability and robustness assessment of these systems, especially when we consider C-NIDS. Thus, the lack of standardized and robust assessments to measure the reliability and accuracy of C-NIDS in attacks detection results in significant uncertainties regarding their ability to correctly identify threats. In this paper, we present an evaluation of a collaborative NIDS that integrates standalone NIDS to share information about detected and mitigated threats, improving overall intrusion detection. The evaluation took place in a virtualized IoT environment and results showed that collaborative NIDS detected DDoS attacks with an accuracy of at least 99.94%, a minimum recall of 99.94%, a precision of 100%, and F1 score of 1 in all evaluated scenarios. Such results highlight the development and adoption of uniform evaluation criteria to ensure the effectiveness and reliability of C-NIDS.