Data Security Challenges in AI-Enabled Medical Device Software

Abstract

The potential of AI to develop innovative applications that can benefit healthcare professionals and patients has created interest, especially in Medical Device Software (MDS) domain. However, the adoption of AI in MDS domain has created several challenges which include: making AI transparent; gaps in clarifying accountability; risk associated with the adaptive nature of AI algorithms; mitigating bias in data; lack of regulatory guidelines specific for AI; and assuring data security. Assuring data security is crucial for AIenabled MDS, as compromising sensitive personal health data can create privacy and ethical concerns and sometimes lead to life threatening issues. In this paper, we discuss the importance of adopting AI in the healthcare domain, the importance of data security in AI-enabled MDS, and the data security challenges that AI has brought to the healthcare industry. Additionally, we consider the reasons for the existence of these challenges. The challenges discussed in this paper are in relation to (1) preventing data breaches; (2) preventing adversarial attacks; (3) preventing cyberattacks; (4) preventing insider threats; (5) lack of skilled and trained staff in data security; and (6) complexity of existing standards and lack of security control implementation details.