Towards Zero-Trust Green Security: Intelligent Interference Management and Rechargeable Jamming Mines in Ultra-Dense Networks

Deploying small cells to densify heterogeneous networks (HetNets) provides a scalable way to meet the growing demand for diverse mobile services. Aggressive frequency reuse in ultra-dense networks (UDNs) further enhances spectral efficiency but intensifies inter-cell interference. With intelligent management, this interference can be harnessed to strengthen network-wide physical layer security (PLS). To this end, and in line with green-security principles, we propose that base stations collaboratively shape interference and deploy rechargeable jamming mines (RJMs) to expand secure zones by injecting targeted jamming around potential eavesdroppers, eliminating the need for dedicated jammers. To realize this, we develop a multi-agent reinforcement learning framework in which each BS independently optimizes beam orientation, transmit power, resource block allocation, and RJM activation under a zero-trust model that treats every location as a potential eavesdropper. A hybrid reward function aligns local actions with network-wide security goals, and state-dependent action masking reduces the policy search space. Extensive results show that the proposed approach substantially enlarges secure-area coverage, scales efficiently in ultra-dense deployments, and establishes a practical, sustainable foundation for collaborative PLS in next-generation wireless networks.