HITL-EdgeSec: A Human-in-the-Loop Autoencoder Framework for Anomaly Detection in 6G Edge Networks

The advent of sixth-generation (6G) technologies will enable ultra-low latency communications, ubiquitous connectivity, and edge intelligence. At the same time, these new networks with their vast threat landscape are susceptible to unprecedented cyberattacks. In this paper, we present HITL-EdgeSec, a human-in-the-Ioop (HITL) anomaly-based intrusion detection system employing unsupervised autoencoders (AEs) in the network edge. With human feedback incorporated at the classification process, HITL-EdgeSec reduces false positives while retaining the ability to detect unknown attacks. We evaluate the framework via a dataset of daily network traffic and simulate human decision-making to estimate the benefit of HITL. Our results demonstrate that combining lightweight unsupervised detection with selective human verification improves detection accuracy and deployability in real-world applications for 6G security scenarios.