ENFOR-SA: End-to-end Cross-layer Transient Fault Injector for Efficient and Accurate DNN Reliability Assessment on Systolic Arrays

2026-01-31

AI Summary

This work addresses the challenge of balancing accuracy and efficiency in evaluating the reliability of systolic-array-based deep neural networks (DNNs) under transient faults. The authors propose an end-to-end cross-layer transient fault injection framework that employs an RTL-level systolic array model exclusively during the fault injection phase, while offloading all other computations to the software layer. This approach demonstrates for the first time that RTL-level accuracy can be preserved without requiring full-cycle, high-overhead RTL simulation. By leveraging a two-stage cross-layer simulation methodology, the framework achieves substantial performance gains: experiments show an average speedup of 569× over full SoC RTL simulation and 2.03× over state-of-the-art cross-layer tools, with only a 6% slowdown compared to pure software-based injection—all while maintaining hardware-level fidelity in fault impact assessment.

Technology Category

• Machine Learning: Hardware-aware ML
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization
• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Recent advances in deep learning have produced highly accurate but increasingly large and complex DNNs, making traditional fault-injection techniques impractical. Accurate fault analysis requires RTL-accurate hardware models. However, this significantly slows evaluation compared with software-only approaches, particularly when combined with expensive HDL instrumentation. In this work, we show that such high-overhead methods are unnecessary for systolic array (SA) architectures and propose ENFOR-SA, an end-to-end framework for DNN transient fault analysis on SAs. Our two-step approach employs cross-layer simulation and uses RTL SA components only during fault injection, with the rest executed at the software level. Experiments on CNNs and Vision Transformers demonstrate that ENFOR-SA achieves RTL-accurate fault injection with only 6% average slowdown compared to software-based injection, while delivering at least two orders of magnitude speedup (average $569\times$) over full-SoC RTL simulation and a $2.03\times$ improvement over a state-of-the-art cross-layer RTL injection tool. ENFOR-SA code is publicly available at https://github.com/rafaabt/ENFOR-SA.

Problem

• transient fault injection
• DNN reliability
• systolic array
• RTL simulation
• fault analysis

Innovation

• systolic array
• transient fault injection
• cross-layer simulation
• DNN reliability
• RTL-accurate