DR4: From Failures to Safety Cases

Small Unmanned Aerial Systems (sUAS) must adhere to strict safety requirements when used in high-stress emergency response situations. However, the increasing use of sUAS raises the risk of major incidents, a good percent of which have been human-made. To tackle this, researchers have been using Fuzz testing to find possible human-prone errors. Fuzz testing is an approach to software testing aimed at identifying system vulnerabilities and defects by injecting random inputs. Human Interaction Fuzzing for small Unmanned Aerial Vehicles (HIFuzz) system [1] is a fuzz testing pipeline developed to identify possible human inputs that could lead to drone mission failures. My project creates safety cases in the form of decision trees to visualize the project's goals, strategies, solutions, HIFuzz test data, and additional information. These safety cases are intended to help understand how to address issues with the HIFuzz system, set goals, and develop solutions.  The ultimate goal of  HIFuzz is to autonomously generate safety cases by taking in data and producing a complete and detailed safety case with all relevant information. Because the end game of the project is to have the safety cases be created autonomously they will need to be reviewed post-rendering by humans on the loop.  Therefore, we created a web-based interactive safety case, featuring the ability to edit text in the nodes, create nodes, delete nodes, and move nodes and the decision tree itself. By implementing my project into the overall HIFuzz project, the human-on-the-loop can visualize the goals and strategies needed to combat issues they will encounter while seeking to improve their overall system.

[1] Chambers, Theodore, Michael Vierhauser, Ankit Agrawal, Michael Murphy, Jason Matthew Brauer, Salil Purandare, Myra B. Cohen, and Jane Cleland-Huang. "HIFuzz: Human Interaction Fuzzing for Small Unmanned Aerial Vehicles." In Proceedings of the CHI Conference on Human Factors in Computing Systems, pp. 1-14. 2024.