A Secure Multiradio Resource Scheme Using Cooperative DRL Agents for Heterogeneous Inter-RAN Slicing Under Hardware Impairments

Network slicing (NS) is an innovative technology that shapes the architecture of sixth-generation (6G) networks, allowing each slice to meet specific Quality of Service (QoS) requirements. Managing radio resources across dense heterogeneous slices with diverse demands presents significant challenges, especially under the zero-touch network (ZTN) paradigm. To address this concern, in the present study, we propose a secure self-optimizing (SO) scheme to manage multiple radio resources (power and bandwidth) across heterogeneous slices on the interslice level in open-RAN (O-RAN). The main goal of the proposed scheme is to maximize spectral efficiency while ensuring a service-level agreement (SLA) for each slice. The problem is formulated as a partially observed Markov decision process (POMDP) and then solved using a cooperative multi-actor-critic (CoMA2C), named SO-CoMA2C. We further enhance the learning process for each agent using long-short-term memory (LSTM) networks. Also, we integrate the advanced encryption standard (AES) into the deep reinforcement learning (DRL) framework to secure communication between the NS environment and agents to ensure secure resource allocation. The proposed approach considers both ideal and nonideal hardware impairments, in addition to the fluctuating traffic loads, thereby enhancing the practical relevance and robustness of the solution. Extensive simulations under various conditions demonstrate the superiority of SO-CoMA2C compared to state-of-the-art benchmarks. Importantly, the integration of AES introduces only minimal overhead, resulting in a 0.13% increase in training time and a 2.3% increase in memory usage.