Dynamic Fuzzing-Based Whole-System Timing Analysis

Summary

This conference paper introduces Fret, a novel dynamic approach to whole-system timing analysis that employs feedback-guided fuzzing to uncover timing-critical dependencies in real-time systems without requiring prior knowledge of inputs or system states. Unlike traditional compositional methods that often overestimate worst-case response times, Fret models system-level interactions including task communication, OS behaviour, and asynchronous events holistically. Implemented using LibAFL and evaluated on FreeRTOS, the method consistently outperforms existing fuzzing strategies and produces complementary artefacts supporting system validation and robust scheduling.

UK applicability

This work addresses embedded systems and real-time computing challenges that are globally relevant rather than geographically specific. UK institutions and industries working on safety-critical systems (automotive, aerospace, industrial control) may find value in the methodology, though applicability depends on whether systems run FreeRTOS or similar real-time operating systems.

Key measures

Worst-case response times (WCRTs), worst-case execution time (WCET) estimates, timing accuracy compared to state-of-the-art fuzzing strategies, inter-task communication delays, task/OS interaction effects, interrupt scheduling effects

Outcomes reported

The study presents Fret, a dynamic whole-system approach for estimating worst-case response times (WCRTs) in real-time systems using feedback-guided fuzzing. Fret identifies timing-critical dependencies and produces actionable artefacts including worst-case inputs, interrupt schedules, and inter-task programme-flow information.

Topic tags