Next-Generation Security: Detecting Suspicious Liquids Through Software Defined Radio Frequency Sensing and Machine Learning

Hazardous liquids, such as nitroglycerin, are now preferred over traditional explosives in modern terrorist attacks. Their inconspicuous nature poses an identification challenge, necessitating urgent large-scale security inspections to prevent terrorism in public spaces. However, conventional liquid detection methods face obstacles in terms of cost, accuracy, and scalability, hindering their widespread use. This article introduces a platform that combines radio frequency (RF) sensing based on software-defined radio (SDR) technology and state-of-the-art machine learning (ML) algorithms to detect and classify suspicious and nonsuspicious liquids. The detection method utilizes fine-grained samples of orthogonal frequency division multiplexing (OFDM) to acquire channel state information (CSI) of the liquids present in the environment at operating frequencies 900 MHz and 2.45 GHz. ML algorithms are employed for classification purposes based on liquids’ dielectric properties, and their effectiveness is evaluated based on accuracy, prediction speed, and training time. The analysis of experimental results demonstrated that our method successfully classified over 95% of both suspicious and nonsuspicious liquids. It also identified more than 97% of suspicious liquid types and classified up to 98% of nonsuspicious liquids. These findings confirm the efficacy of the proposed system. The software-defined RF sensing system is versatile, portable, adaptable, and cost-efficient.