This paper introduces a new method for detecting architectural drift in safety-critical firmware, specifically for systems compliant with ISO 26262 standards. The approach utilizes hardware-assisted execution traces to model runtime behavior as message exchanges between firmware components. These runtime models are then compared against design-time sequence diagrams to identify discrepancies. A constrained LLM is employed to generate human-readable reports of these differences, aiding expert review and reducing manual analysis effort. Evaluation in an industrial firmware context demonstrated strong agreement with expert references and positive feedback from practitioners on the usefulness of the generated reports. AI
IMPACT This research could improve the reliability and safety of critical firmware by providing better tools for detecting design inconsistencies.
RANK_REASON This is a research paper detailing a new methodology for firmware analysis. [lever_c_demoted from research: ic=1 ai=0.7]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
