CORA: Teacher-Guided Causal O-RAN Adaptation for Jamming Detection

| Source: arXiv AI

Tags: CORA, O-RAN, 5G, Jamming Detection, PPO, DQN

CORA uses a retrospective teacher detector to train a causal, label-free jamming detector for 5G O-RAN, raising mean recall from 0.687 to 0.771 across ten testbed configurations.

Details

In mission-critical networks a missed jamming event costs more than a false alarm, but retrospective KPI detectors that work without labels need future context and cannot run in streaming mode. CORA uses such a retrospective detector offline as a teacher, turning its verdicts on unlabeled historical logs into a recall-weighted reward. An RL controller learns to tune sensitivity, persistence and grace of a causal detector, and the teacher is removed at deployment. The design maps to the Non-Real-Time and Near-Real-Time RICs in O-RAN. Under leave-one-out transfer across ten 5G O-RAN testbed setups, a PPO controller raised mean recall from 0.687 to 0.771, while a DQN variant matched the best static F2-score with 61% fewer false alarms.