Recent advancements in machine learning have enabled market innovators to build insights from IoT sensors in the wild. These insights can be used to solve complex real-world challenges. The lack of security in typical vision-based IoT solutions is especially concerning, as they are typically responsible for managing sensitive data (PID, CCTV) or critical systems (cars, machinery). Security is rarely the first thought for developers of new types of solutions, but making systems secure after the fact is difficult since a holistic approach is required. Exacerbating this challenge of achieving end-to-end security, the development and deployment of IoT systems often involves multiple handovers of responsibility, which can make achieving end-to-end security difficult. And, due to the complexity and diversity of these systems, security bodies have been unable to prescribe “silver bullet” solutions. Based on first-hand experience, this presentation will provide insights to help decision-makers better understand key challenges and potential solutions for providing secure vision-based IoT systems.