Parul Dubey
Computer Science and Engineering Department, Symbiosis Institute of Technology, Nagpur Campus, Symbiosis International (Deemed University), Pune, Maharashtra, India.

Pushkar Dubey
Department of Management, Pandit Sundarlal Sharma (Open) University, Bilaspur, Chhattisgarh, India.

Gagandeep Kaur
Computer Science and Engineering Department, Symbiosis Institute of Technology, Nagpur Campus, Symbiosis International (Deemed University), Pune, Maharashtra, India.

DOI https://doi.org/10.33889/IJMEMS.2026.11.3.055

  ;

Abstract

Massive multiple-input multiple-output (mMIMO) systems are the backbone of modern-day wireless communication due to their potential to utilize spectrum efficiently and increase network capacity. Secure and optimal beam selection is key to interference and security challenges in dense urban areas, especially with the arrival of 5G and beyond. Classical solutions such as exhaustive beam search or even statistical models have high computational complexity and are not so flexible to dynamic situations. This study generated simulated data based on the realistic distribution of the users and the phenomena of the terrestrial. The data set records common metrics like the user locations, channel states, and beamforming settings. We propose a deep-learning framework that predicts top-K transmit–receive beam pairs using only receiver location and then enforces physical-layer security (PLS) by filtering out pairs that violate an eavesdropper-power threshold. On simulated DeepMIMO-inspired scenes, our model attains Top-1/Top-5/Top-10 accuracies of 69.51%/85.32%/92.43%, cuts beam-search overhead by 92.19%, and reduces mean execution time to 95 ms. With security constraints, it approaches achievable bounds for Probability of Successful Detection (PSD)/ Probability of Secure Signal Detection (PSSD) and reduces estimated eavesdropping probability from 15.6% to 5.2%, while improving secrecy capacity and Bit Error Rate (BER). The novelty is a security-constrained beam selection loop integrated directly into initial access, requiring low CSI and remaining deployable within 5G NR procedures.

Keywords- Reference signal received power (RSRP), Security constraints, Top-K beam selection, Massive multiple-input-multiple-output (mMIMO).

Citation

Dubey, P., Dubey, P., & Kaur, G. (2026). Towards Intelligent and Secure Beam Alignment: A Deep Learning Perspective for 5G. International Journal of Mathematical, Engineering and Management Sciences, 11(3), 1346-1369. https://doi.org/10.33889/IJMEMS.2026.11.3.055.