Information Superhighway for Neurological Disorders: Security Vulnerabilities and their mitigation strategies

Artificial Intelligence (AI) is making great strides in the healthcare sector. Its prowess is being increasingly utilized to develop powerful detection & classification models for neurological disorders. Cost efficient AI models are the answer to the otherwise costly approaches of disorder detection through brain scans, behavioral analysis, etc. For an end user to realize the potential of these AI models end to end the architecture of an Information Superhighway in the form of AI pipeline was presented. The AI pipeline enabled capturing the multimodal data (image, video, etc.) of a neurologically affected subject from a smartphone of a laptop webcam, and send this through a secure information channel to a cloud, where it was managed using various Big Data technologies. Further, the DL models were used for generating different types of recommendations, which were sent through a different secure communication channel back to the mobile or laptop. The research work through which AI pipeline was proposed was quite novel. However, we feel that in healthcare systems, data security is a major concern. Hence, in this paper, we discuss security vulnerabilities of the AI pipeline, highlighting different types of security attacks that are possible at different points in the AI pipeline. We also discuss why we feel the particular attack is possible and ways to mitigate the effects of such attacks.