Securing Bio-FET Interfaces in IoBNT Systems using Deep Learning Techniques

Several recent studies to address security vulnerabilities of Internet of Bio-Nano Things (IoBNT) are underway to streamline its wider adoption in e-healthcare. IoBNT systems comprise of in-vivo nano-networks connected to external networks through bio-cyber interfacing for remote drug delivery and healthcare management. Biofield-effect transistors (BioFET) based interfaces have shown promising results in realizing practical IoBNT implementation. BioFETs offer an economical and scalable solution to connect molecular in-body drug delivery networks, actuated by external monitoring stations. Ensuring BioFET operational safety against cyber threats is vital to increasing patient confidence and complying with strict data protection and security governance required by e-healthcare providers. The present work utilizes deep learning technology to classify biochemical signals originating from BioFET sensors, discriminating between normal and anomalous behaviour. During validation several individual and ensemble DL structures are applied on synthesized dataset to determine optimal anomaly classifier. The results demonstrate that a hybrid convolutional and longshort term network ensemble offers relatively high efficiency (92%) and lower latency in comparison with other DL models.