SAFER: Safety Assurances for Emergent Behavior.

Emergent behavior haunts the reliability and safety of complex software systems. Such behavior consists of unexpected operations that can arise at runtime and lead the system to potentially unsafe conditions. The SAFER framework allows users to model systems using Periodic State Machines and automatically generates code from a specification. It then creates a model that learns safe execution based on execution traces from the code. The applicability of SAFER is demonstrated through four case studies: Producer-Consumer, Collision Avoidance, Integer Overflow, and Stack Overflow. These case studies showcase the framework’s ability to deal with emergent behaviors for different classes of applications.