Self-taught algorithms to improve wireless communications

Wireless networks and the Internet of Things (IoT) are used for monitoring, data collection and analysis in numerous research and applied fields, such as autonomous driving, interconnection of medical systems or industrial equipment, building security, etc.

However, these networks often lose their efficiency in the face of a security attack, leading to miscommunication between devices/sensors connected to that network and, ultimately, to a loss of data.

To address this, researchers at the Universitat Politècnica de València (UPV), jointly with researchers of the Islamia College Peshawar, CCIS Prince Sultan University and of the Prince Sattam Bin Abdulaziz University, have developed an algorithm for IoT networks to improve data exchange without additional costs and without losing information between sensors.

This algorithm uses the machine learning technique and, through intelligent methods, ensures low-latency data routing. Moreover, unlike most existing solutions today, it increases the robustness and security of the mobile network: it protects devices and data from security attacks, allows for more secure device authentication of devices and improves privacy.

With these algorithms, networks adapt to situations through automated agents that continuously learn, share information, and cooperate to achieve an optimised distribution of data in real-time.

"Our algorithm reduces threats in the presence of anonymous devices and increases the reliability of the IoT-enabled communication system thanks to the machine learning techniques it incorporates. It provides intelligent methods to detect the coverage area and efficiently distribute the power load among mobile devices, ensuring efficient communication between devices and optimal network performance," explains Jaime Lloret, a researcher at the Department of Communications and director of the IGIIC Institute at the UPV.

Testing in intelligent buildings

The algorithm developed by the authors was tested in intelligent buildings to interconnect various operations and for security surveillance using IoT mobile devices and different sensors.

This algorithm can collect real-time data from the intelligent building and transmit it to network applications for further analysis and appropriate actions. According to tests carried out, it improves the information packet delivery rate by 17 %, reduces data delay by 22 %, and reduces energy consumption by 24 % and computational complexity by 17 %.

"In the future, we aim to use the deep learning model and the real-time dataset to train the network nodes to deal with network anomalies," concludes Jaime Lloret.

Reference

Haseeb, K.; Rehman, A.; Saba, T.; Bahaj, S.A.; Lloret, J. Device-to-Device (D2D) Multi-Criteria Learning Algorithm Using Secured Sensors. Sensors 2022, 22, 2115. https://doi.org/10.3390/s22062115