Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Spatiotemporal-Aware Bit-Flip Injection on DNN-based Advanced Driver Assistance Systems

Taibiao Zhao, Xiang Zhang, Mingxuan Sun, Ruyi Ding, Xugui Zhou

Abstract

Modern advanced driver assistance systems (ADAS) rely on deep neural networks (DNNs) for perception and planning. Since DNNs' parameters reside in DRAM during inference, bit flips caused by cosmic radiation or low-voltage operation may corrupt DNN computations, distort driving decisions, and lead to real-world incidents. This paper presents a SpatioTemporal-Aware Fault Injection (STAFI) framework to locate critical fault sites in DNNs for ADAS efficiently. Spatially, we propose a Progressive Metric-guided Bit Search (PMBS) that efficiently identifies critical network weight bits whose corruption causes the largest deviations in driving behavior (e.g., unintended acceleration or steering). Furthermore, we develop a Critical Fault Time Identification (CFTI) mechanism that determines when to trigger these faults, taking into account the context of real-time systems and environmental states, to maximize the safety impact. Experiments on DNNs for a production ADAS demonstrate that STAFI uncovers 29.56x more hazard-inducing critical faults than the strongest baseline.

Spatiotemporal-Aware Bit-Flip Injection on DNN-based Advanced Driver Assistance Systems

Abstract

Modern advanced driver assistance systems (ADAS) rely on deep neural networks (DNNs) for perception and planning. Since DNNs' parameters reside in DRAM during inference, bit flips caused by cosmic radiation or low-voltage operation may corrupt DNN computations, distort driving decisions, and lead to real-world incidents. This paper presents a SpatioTemporal-Aware Fault Injection (STAFI) framework to locate critical fault sites in DNNs for ADAS efficiently. Spatially, we propose a Progressive Metric-guided Bit Search (PMBS) that efficiently identifies critical network weight bits whose corruption causes the largest deviations in driving behavior (e.g., unintended acceleration or steering). Furthermore, we develop a Critical Fault Time Identification (CFTI) mechanism that determines when to trigger these faults, taking into account the context of real-time systems and environmental states, to maximize the safety impact. Experiments on DNNs for a production ADAS demonstrate that STAFI uncovers 29.56x more hazard-inducing critical faults than the strongest baseline.

Paper Structure

This paper contains 19 sections, 2 equations, 7 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Background and Related Works
  3. Methodology
  4. Threat Model
  5. Progressive Metric-guided Bit Search
  6. Taylor-guided weight selection
  7. Progressive Metric-guided Bit Search
  8. Critical Fault Time Identification
  9. Experimental Evaluation
  10. Driving Scenarios
  11. Fault Injection
  12. Baselines
  13. Results
  14. Spatial Bit Selection Evaluation
  15. End-to-end Evaluation
  16. ...and 4 more sections

Figures (7)

  • Figure 1: ADAS Overview.
  • Figure 2: Spatiotemporal-Aware Bit Flips Fault Injection Overview. 1: Progressive Metric-Guided Bit Search (PMBS) identifies critical weight bits by combining Taylor-guided importance and metric (Section \ref{['section:PMBS']}). 2: CFTI selects the fault injection time candidates using safety system-context rules (Section \ref{['section:CFTI']}). 3: The corrupted model is evaluated in a closed-loop autonomous driving simulation platform. 4: Bit flips induce hazardous behaviors such as acceleration, lane departure, and collisions.
  • Figure 3: Simulation snapshots.
  • Figure 4: Scenarios. EV: Ego Vehicle. LV: the Lead Vehicle.
  • Figure 5: Flipped bits heatmap.
  • ...and 2 more figures