Journal of Intelligent Systems and Internet of Things

In an increasingly fast-paced world of 6G-IoT networks, optimal beamforming techniques will be effective in improving strength, latency, and quality of service delivery in the networks. This work presents a new paradigm in beamforming optimization, especially in tackling dynamic environments and high computational costs in existing approaches. The problems of long training times with traditional methods, along with threats in security make them out rightly less applicable for real time applications. The data is collected from 6G IoT networks then, Laplacian Eigenmaps is used for feature extraction and modelling in time and applied for dimensionality reduction, ConvMarkov is used for model development RC4 encryption secures data exchange, while blockchain supports data logging and promotes transparency. This is a combination of deep learning techniques and advanced encryption methods, which will lead to a wide boost in beamforming efficiency, flexibility, and security. This study achieved the beamforming optimization achieved 97% accuracy with significant gain improvements, as indicated by an ROC curve (AUC = 0.9970) and precision-recall curve. The training loss stabilized below 0.01, while the validation loss fluctuated above 0.1, suggesting minor overfitting. The main achievements converge on proving improvements in optimization under real time conditions in a network, besides integrity and privacy of data. These become great merits into a strong solution for future 6G.