STAFF: Stateful Taint-Assisted Full-system Firmware Fuzzing

2025-09-22

AI Summary

Existing firmware fuzzing approaches suffer from blind spots when testing embedded Linux devices, primarily due to deep inter-process state dependencies among multiple daemons—dependencies that conventional single-process, coverage-guided fuzzers cannot model or traverse. Method: This paper proposes the first firmware-level fuzzing framework with full-system state awareness. It integrates user-driven multi-request recording, cross-process dependency inference, whole-system taint analysis, and protocol-aware taint-guided mutation. Crucially, it introduces a novel multi-stage forkserver snapshotting mechanism and protocol-state checkpoints to precisely model and efficiently explore persistent internal states and inter-daemon interactions. Contribution/Results: Evaluated on 15 real-world firmware images, the framework discovered 42 reproducible multi-request, multi-daemon vulnerabilities—significantly outperforming state-of-the-art single-process, coverage-guided fuzzers in both quantity and exploitability.

Technology Category

• Machine Learning: Hardware-aware ML
• Search and Optimization: Distributed Search
• Multiagent Systems: Other Foundations of Multi Agent Systems

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing
• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Modern embedded Linux devices, such as routers, IP cameras, and IoT gateways, rely on complex software stacks where numerous daemons interact to provide services. Testing these devices is crucial from a security perspective since vendors often use custom closed- or open-source software without documenting releases and patches. Recent coverage-guided fuzzing solutions primarily test individual processes, ignoring deep dependencies between daemons and their persistent internal state. This article presents STAFF, a firmware fuzzing framework for discovering bugs in Linux-based firmware built around three key ideas: (a) user-driven multi-request recording, which monitors user interactions with emulated firmware to capture request sequences involving application-layer protocols (e.g., HTTP); (b) intra- and inter-process dependency detection, which uses whole-system taint analysis to track how input bytes influence user-space states, including files, sockets, and memory areas; (c) protocol-aware taint-guided fuzzing, which applies mutations to request sequences based on identified dependencies, exploiting multi-staged forkservers to efficiently checkpoint protocol states. When evaluating STAFF on 15 Linux-based firmware targets, it identifies 42 bugs involving multiple network requests and different firmware daemons, significantly outperforming existing state-of-the-art fuzzing solutions in both the number and reproducibility of discovered bugs.

Problem

• Testing complex embedded Linux firmware with interdependent daemons
• Overcoming limitations of process-focused fuzzing ignoring persistent state
• Discovering bugs requiring multiple requests across different firmware components

Innovation

• User-driven multi-request recording for protocol sequences
• Whole-system taint analysis for dependency detection
• Protocol-aware taint-guided fuzzing with multi-staged forkservers