Stateful Embedded Fuzzing with Peripheral-Accurate SystemC Virtual Prototypes Conference: 63rd ACM/IEEE Design Automation Conference; July 26–29, 2026; Long Beach, CA, USA 63rd ACM/IEEE Design Automation Conference (DAC ’26), July 26–29, 2026, Long Beach, CA, USA DOI: 10.1145/3770743.3804000 ISBN: 979-8-4007-2254-7/2026/07 CCS: Software and its engineering Software testing and debugging CCS: Computing methodologies Discrete-event simulation
Chiara Ghinami, Igor Pontes Tresolavy, Luis Seibt, Nils Bosbach, Rainer Leupers · 2026
Read Stateful Embedded Fuzzing with Peripheral-Accurate SystemC Virtual Prototypes Conference: 63rd ACM/IEEE Design Automation Conference; July 26–29, 2026; Long Beach, CA, USA 63rd ACM/IEEE Design Automation Conference (DAC ’26), July 26–29, 2026, Long Beach, CA, USA DOI: 10.1145/3770743.3804000 ISBN: 979-8-4007-2254-7/2026/07 CCS: Software and its engineering Software testing and debugging CCS: Computing methodologies Discrete-event simulation by Chiara Ghinami, Igor Pontes Tresolavy, Luis Seibt, Nils Bosbach, Rainer Leupers (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-19824.txt.
Introduction
Ghinami and four colleagues present a framework that joins AFL++ with a stateful SystemC-TLM virtual prototype to fuzz embedded software true to life. Fuzzers work well for conventional software but struggle in embedded settings, where peripheral behavior matters. Older routes use fast user-mode simulators that drop peripheral realism, or full-system simulators with manual setup that limit large-scale use. Fuzzer-generated inputs are fed straight into the virtual prototype.
Catalog record
| CREATOR | Chiara Ghinami, Igor Pontes Tresolavy, Luis Seibt, Nils Bosbach, Rainer Leupers |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 30977 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.19824
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Stateful Embedded Fuzzing with Peripheral-Accurate SystemC Virtual Prototypes Conference: 63rd ACM/IEEE Design Automation Conference; July 26–29, 2026; Long Beach, CA, USA 63rd ACM/IEEE Design Automation Conference (DAC ’26), July 26–29, 2026, Long Beach, CA, USA DOI: 10.1145/3770743.3804000 ISBN: 979-8-4007-2254-7/2026/07 CCS: Software and its engineering Software testing and debugging CCS: Computing methodologies Discrete-event simulation Chiara Ghinami, Igor Pontes Tresolavy, Luis Seibt, Nils Bosbach, Rainer Leupers Abstract . The increasing complexity of embedded software has made comprehensive manual testing impractical, motivating the use of automated techniques such as fuzzing. Coverage-guided fuzzers like AFL++ have shown strong results for conventional software but remain challenging to apply effectively in embedded contexts, where peripheral behaviors play critical roles. Existing approaches either use fast user-mode simulators, sacrificing peripheral realism, or rely on full-system simulators with manual instrumentation, limiting applicability to large-scale software. In this work, we present a novel framework that integrates AFL++ with a stateful SystemC-TLM virtual prototype to enable realistic fuzzing of embedded software. Fuzzer-generated inputs are injected directly into peripheral models, allowing peripherals to trigger natural side effects such as interrupts and FIFO updates. By integrating fuzzing with full-system simulation, our framework advances the effectiveness of pre-silicon testing for embedded systems. Results on embedded workloads show that our approach eliminates false positives while maintaining comparable code coverage and execution performance as state-of-the-art tools. AFL American Fuzzy Lop SoC System-On-Chip PUT Program Under Test MMIO Memory-Mapped Input/Output VCML Virtual Components Modelling Library ISS Instruction Set Simulator FSS Full-System Simulator VP Virtual Prototype TLM Transaction-Level Modeling DBT Dynamic Binary Translation AVP Arm Virtual Platform NVIC Nested Vector Interrupt Controller FTL Fast Translation Library OS Operating System MCU Microcontroller ADC Analog-to-Digital Converter RTC Real-Time Clock HAL Hardware Abstraction Layer ISR Interrupt Service Routine ECU Electronic Control Unit RTOS Real-Time Operating System † † cc-license: by 1. Introduction In the past decade, the complexity of embedded software has increased significantly to cope with real-world problems, making manual software testing practically impossible. Fuzzing, a popular testing technique 15 ; 28 , feeds random inputs to the put ( put ) and, being fast and lightweight, avoids issues, such as the state explosion, that affect symbolic execution 15 . Figure 1 . Simulation-based fuzz testing. AFL++ 9 , a community-maintained fork of the AFL fuzzer 23 , is widely used in industry and research for coverage-guided testing. Recently, significant research has focused on enabling embedded software fuzzing by integrating AFL++ with simulators. As shown in Fig. 1 , the fuzzer takes a seed from the user, mutates it, and gives it as input to the put that runs on the simulator. During the execution, the simulator gathers code coverage information. Based on that, the fuzzer chooses to discard the input or further use it and add it to the corpus. Functional simulators such as QEMU 1 offer fast software execution along with debugging capabilities, making them ideal for early-stage software testing and validation. For this purpose, both user-mode and full-system simulators are commonly used. User-mode simulators do not simulate the target os ( os ) and peripherals, which limits them to user-space applications. Instead, they forward the system calls of the target software to the host os . In contrast, full-system simulators, or vp , model the entire hardware platform, including peripherals, enabling the simulation of the complete target software stack. They provide higher execution fidelity at the cost of reduced performance. Many research efforts favor user-mode simulators for embedded fuzzing 5 ; 6 ; 18 ; 27 . These approaches learn which register values influence the control flow and inject them accordingly. While effective in increasing execution speed, frameworks such as Fuzzware 18 and P2IM 6 fail to model peripheral causality (interrupts, FIFO and register updates), which can result in unrealistic execution traces and high false positives count. Other works have employed full-system simulators for fuzzing 11 ; 22 . However, these approaches require the modification of peripheral models or manual put instrumentation, limiting their broader adoption. Unlike prior works that rely on program analysis to infer register values, our framework integrates AFL++ with a full-system SystemC-TLM vp and introduces a novel injection mechanism that (1) intercepts mmio ( mmio ) accesses and (2) injects fuzzer-generated values directly into virtual peripheral models on selected mmio events, allowing peripherals to produce realistic side effects, without any modifications required to either the peripheral models or the fuzzed source code. Motivated by recent demonstrations of attacks that inject malicious data through communication interfaces (e.g., CAN and UART) 12 ; 17 ; 21 , we target software that interacts directly with soc ( soc ) peripherals, such as bare-metal applications, rtos ( rtos ) device drivers, and hal to test whether external data can trigger software faults. In several experiments, we show that our framework outperforms prior work in execution fidelity, when evaluated on embedded workloads that exercise a wide range of peripherals. With this work, we bridge the gap between fuzz testing and full-system simulation, enabling more realistic and effective testing of embedded systems. 2. Background In this section, we provide an overview of vp and fuzzing, followed by a summary of existing approaches that leverage simulators for embedded fuzzing. 2.1. Virtual Prototypes A vp is a software model of a hardware system that executes the same software as the physical hardware, producing functionally equivalent outputs. To showcase our approach, we utilize SIM-A, a proprietary ARM-based VP developed by MachineWare 14 . SIM-A is a functional simulator, meaning it focuses on the logical correctness of computation. Unlike cycle-accurate simulators, which capture microarchitectural details, functional simulators assume a fixed number of cycles per instruction, ignoring pipeline stages, stalls, or cache effects, to maximize the execution speed. SIM-A is implemented using SystemC- tlm ( tlm ) 20 , a C++-based modeling framework that abstracts low-level hardware behavior and provides better modularity and customizability compared to traditional C-based functional simulators such as QEMU 1 . Thanks to the SystemC-TLM standardized interface, that provides a common way to connect components, our approach is not tied to SIM-A specifically and can be applied to any SystemC-TLM-based VP with minimal adaptation. For peripheral models, we rely on the open-source Virtual Components Modeling Library (VCML) 13 . VCML provides a rich set of pre-defined peripheral components that can be directly connected to a SystemC CPU model, enabling the simulation of a broad range of SoC configurations. Leveraging these existing peripheral models, originally developed for general VP use cases, allowed us to significantly reduce the implementation effort while ensuring realistic device behavior. 2.2. Fuzzing Fuzzing is a software-testing technique that aims to uncover bugs in the put by providing unexpected or malformed inputs, often triggering corner-case behaviors. The most adopted taxonomy classifies fuzzers based on how much information is used from the program execution 16 . Black-box fuzzers operate without any feedback from program execution. At the opposite end, white-box fuzzers select the input based on extensive program analysis, such as symbolic execution 10 . In between, grey-box fuzzers use lightweight execution feedback to guide input generation, improving coverage and bug detection over black-box approaches, while avoiding the high overhead of white-box analysis 2 ; 4 ; 9 ; 8 . This makes grey-box fuzzers well-suited for large-scale and