Originally published by the High Confidence Software and Systems Conference.
High Confidence Software and Systems Conference, Annapolis, MD, May 11-12, 2026
Author:
Dr. Chris Wright
Poster:
Abstract:
High-confidence assurance of software-intensive and cyber-physical systems depends not only on the availability of advanced analysis techniques, but on the ability to translate their outputs into structured and traceable artifacts that support assurance reasoning and decisionmaking. In practice, firmware and binary analysis tools produce fragmented results that remain difficult to integrate into coherent assurance workflows, limiting their usefulness beyond expert reverse engineering.
This talk presents REAFFIRM, a framework designed to bridge the gap between low level firmware analysis outputs and assurance-relevant artifacts. Given a firmware image or binary, REAFFIRM orchestrates a range of static, dynamic, and semantic analyses, including capability analysis, similarity and labeling techniques, symbolic execution for hardware interaction discovery, and firmware bill of materials extraction, to generate structured evidence suitable for downstream assurance tasks.
Rather than focusing on a single analysis technique, REAFFIRM emphasizes the organization and synthesis of evidence to support multiple assurance perspectives. Analysis results are persisted as structured artifacts that can be queried to answer questions such as which components interact with hardware, which subsystems warrant further validation, and what evidence supports specific risk or capability claims. This enables repeatable, goal-directed assurance workflows that move beyond ad hoc scripting and manual interpretation. REAFFIRM also integrates selective use of machine learning and large language models to assist analysts in interpreting results and translating technical findings into role-specific reports for red teams, blue teams, and executive stakeholders. These AI techniques are used to support human understanding and communication while preserving traceability to underlying analysis evidence. The framework further supports targeted validation activities by enabling analysts to extract components for fuzzing and partially automate test setup, while explicitly leaving semantic input modeling to human expertise.
This work demonstrates how organizing analysis outputs into assurance artifacts can reduce analyst burden while preserving rigor, transparency, and defensibility, and illustrates an evidence-driven approach to firmware assurance that augments human workflows without obscuring uncertainty or judgment. This talk is directly relevant to the High Confidence Software and Systems Conference and addresses the AI as an Enabler theme by showing how AI-assisted techniques can support structured, evidence-driven firmware assurance workflows while preserving human judgment, traceability, and rigor.