Unified DRSN-Deep Cross-LSTM Framework for Trust Management in 6G-IoT Networks

In the rapidly growing area of wireless communication, the upcoming 6G networks are predicted to transform the method we interconnect, presenting supreme speed, least latency, and consistent connectivity. The introduction of innovative technologies in 6G might form novel exposures, which cybercriminals might develop. Furthermore, the incorporation of huge IoT with 6G poses extra hazards, as it considerably increases the surface of an attack, thus complicating the protection of a previously complex and extensive network structure. Among this development, the supreme concern remains the privacy and security of data traversing these systems. The fast development of 6G-enabled Internet of Things (IoT) systems has converted communication and made it probable for healthcare, industrial automation driverless cars, and smart cities to all have perfect connectivity. However, the rising diversity and complexity of 6G-IoT structures present dangerous cybersecurity problems, leaving these systems open to a sort of attacks like spoofing, malware dissemination, and Distributed Denial-of-Service (DDoS). It is because conventional intrusion detection systems (IDS) cannot alter to altering attack patterns, they are unable to effectively combat these threats. This paper develops a new Deep Residual Shrinkage Network with a Deep cross-network (DRSN-DCLSTM) model for Trust and Reputation Management on 6G-IoT Networks. The main intention of the DRSN-DCLSTM system is to enhance security in the 6G-IoT networks through the recognition of intrusions. The proposed DRSN-DCLSTM model exploits classification performance while decreasing the cost of computational by employing a selection of feature and hyperparameter tuning. For data normalization, the DRSN-DCLSTM technique initially employs a Quantile Transformer which achieves mapping of input data to a uniform or normal distribution by the usage of quantiles. Besides, the soft-thresholding process in DRSN allocates an adaptive threshold to every individual sample, which efficiently removes redundant features and improves the efficiency of training. In the classification stage, a deep cross-LSTM network has been employed which recognizes the intrusions proficiently in 6G-IoT systems. Eventually, a superior Sophia, an effective second-order optimizer utilizing Hessian approximations is employed in order to upgrade the model parameters. Training and evaluation are led on the RT-IoT database, showing the efficiency of the method. The experimental outcomes establish that the proposed DRSN-DCLSTM model outperforms individual machine learning techniques, attaining improved detection accuracy.