Επέκταση λογισμικού unit testing για τη διεξαγωγή fuzz testing δοκιμών σε ενσωματωμένο λογισμικό

Δάσκαλος, Άγγελος - Χρήστος

Extension of unit testing software for the fuzz testing on an embedded unit software

Master Thesis

Author

Δάσκαλος, Άγγελος - Χρήστος

Date

2025-11

Abstract

Embedded systems are often designed using serial protocols controlled by complex state machines, but these critical mechanisms often remain insufficiently tested. This paper proposes a systematic methodology for enhancing and validating the functionality of such protocols using stub-based unit testing and coverage-guided fuzzing techniques. Conducting a fuzzing campaign on an embedded system is a particularly complex process, as it requires modifying the fuzzer implementation (e.g., fuzzing harness) for each piece of software under test. In addition, fuzzing campaigns often need to be run in a simulation environment or under constraints imposed by the hardware and peripherals. In this work, we aim to automatically generate the appropriate software and data so that embedded system projects can immediately reap the benefits of fuzz testing, with checks at the software module/object file level that can be efficiently executed on third-party systems (even those with a different architecture from that of the embedded system). For the experimental testing of the methodology, firmware was developed for the STM32 microcontroller that drives a servo motor where the peripherals are replaced with stubs, allowing the isolation and verification of the authentication, menu navigation, checksum verification, and command cancellation. Using CMocka, test units are developed that reveal logical malfunctions and inconsistencies in the state machine prior to firmware integration. Subsequently, LibFuzzer, in combination with AddressSanitizer and UBSanitizer, feeds millions of malformed byte packets and edge bytes to the UART protocol analyzer, revealing configuration memory issues and failures in the protocol. All verification steps are orchestrated within a version-controlled workflow on GitHub and encapsulated in Docker containers, ensuring reproducible builds and test environments. Our combined methodology yields significant coverage gains across both protocol parsing and servo-control state transitions, exposing subtle vulnerabilities such as checksum bypasses, improper state resets, and cancellation race conditions. By embedding stub-driven unit testing and fuzzing into a Docker-enabled GitHub pipeline, we present a reproducible framework for elevating reliability in resource-constrained embedded devices.

Postgraduate Studies Programme

Κυβερνοασφάλεια και Επιστήμη Δεδομένων

Department

Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα Πληροφορικής

Number of pages

158

Language

Greek

URI

https://dione.lib.unipi.gr/xmlui/handle/unipi/18549

Collections

Τμήμα Πληροφορικής

Show full item record

Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού 3.0 Ελλάδα