Deep learning-based security schemes for implantable medical devices

Abstract Deep learning is a subset of machine learning, which learns from the inherent patterns in the data for solving a diverse set of problems such as recognition, classification, and segmentation. It is a neural network-based, biologically inspired model, which has benefitted health, transport, energy, and public safety sectors in diverse ways. It has enabled new potential innovations in these domains, including data analytics, security, treatment, and diagnostics. Intelligent healthcare enables medical specialists to remotely monitor patients, thereby leading to an increase in the popularity of this field in recent years. Doctors are able to provide a better quality of treatment to their patients through a variety of implanted medical devices. The addition of communication ability enables such devices to talk with one another and to the Internet, which leads to the concept of the Internet of Things applied for medical devices. Such devices now have 802.11x or LTE chips on, with the goal that they can converse with one another, in addition to the conventional jobs of sensing and actuating. However, on the other end, the addition of wireless connectivity now makes these devices too prone to be hacked, leading sometimes to lethal events for patients if they are not mitigated. This chapter focuses on how deep learning can be utilized to make these devices more secure while addressing the tradeoffs related to constrained computations, and energy available on such devices.