Cyber-physical systems connect the real, physical world to computation, for example in the domain of autonomous vehicles. Because of the real world instantiation and potential risks, safety concerns are paramount. Safe reinforcement learning refers to machine learning that incorporates considerations of real world safety. This CAREER project focuses on enhancing the security of cyber-physical systems that are being designed using the current state of the art safe reinforcement learning methods. In general, safe learning systems focus on performance under safety constraints, however, they remain vulnerable to attacks during operation or training. Achieving safe and secure reinforcement learning protects users from systems and systems from attack. This project will develop innovations that focus on achieving these goals using precise specifications expressed in Signal Temporal Logic (STL) for studying both functional and timing vulnerabilities in these systems and eventually designing mitigation strategies. Evaluation will leverage the CARLA (CAR Learning to Act) simulator for autonomous driving research and real-world autonomous car testbeds to validate security measures, ensuring resilient CPS deployment in complex and adversarial conditions. Overall, this CAREER project will lead to improvement in the security of Cyber-Physical Systems (CPS) such as autonomous vehicles that utilize reinforcement learning in their operation. The project will lead to the discovery of potential security risks that target the learning process and real-time operation of the vehicle. The project will develop real-time detection and diagnostic tools and methods that will harden the vehicle against these risks – especially those associated with the learning and training process. By addressing these security gaps, this research will help ensure the cyber physical systems operate reliably in real-world environments, ultimately improving safety in transportation, robotics, and other critical applications. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.