Under pressure: Effect of a ransomware and a screen failure on trust and driving performance in an automated car simulation

Payre, William; Perelló-March, Jaume; Birrell, Stewart · 2023 · Crossref

DOI: 10.3389/fpsyg.2023.1078723

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates how cybersecurity vulnerabilities and system failures in connected and automated vehicles (CAVs) impact driver trust and manual driving performance. Motivated by the increasing risk of cyberattacks and the potential for "silent" hardware failures to compromise safety without user notification, the research addresses a gap in understanding how drivers react to such events when no explicit takeover request is issued. The authors hypothesized that an explicit failure, simulated as a ransomware attack, would negatively affect trust and driving performance more severely than a silent failure, characterized by the absence of turn signal indicators on the in-vehicle display. The experiment utilized a high-fidelity driving simulator with 38 participants engaging in conditionally automated driving (SAE Level 3). Participants performed a visually demanding non-driving related task (NDRT), a word search, while the vehicle operated in automated mode. The study employed a within-subjects design with three conditions: a control condition, a silent failure condition, and an explicit ransomware condition. In the explicit condition, a ransomware message demanding Bitcoin appeared on the touchscreen during an overtaking maneuver. Crucially, drivers were not prompted to take control; they could voluntarily resume manual driving at any time. Trust was measured using the Trust in Automation Scale and Situational Trust Scale, while driving performance was assessed via lateral control (steering wheel angle standard deviation), longitudinal control (speed homogeneity), and manual takeover frequency. Results indicated that objective trust decreased following both types of failures. Drivers who resumed manual control reported significantly lower trust scores than those who remained in automated mode, particularly after the explicit ransomware event. Behaviorally, significantly fewer drivers resumed their NDRT after the explicit failure compared to the silent failure, suggesting heightened vigilance or distrust. Regarding driving performance, lateral control was compromised for drivers who took over control after either failure, evidenced by greater steering wheel angle variability. However, longitudinal control, measured by speed homogeneity, was smoother during manual driving than during automated driving. One crash occurred in the explicit condition shortly after a driver took control. The findings suggest that connectivity and cybersecurity failures significantly erode trust in automation, with explicit threats like ransomware causing more pronounced behavioral changes than silent hardware faults. The study posits that engagement in NDRTs can serve as a surrogate measure for trust, as disengagement indicates a loss of confidence in the system. Furthermore, the results highlight that while drivers may attempt to regain control after failures, their manual performance, particularly lateral stability, may be degraded. This underscores the importance of robust cybersecurity and transparent failure communication in maintaining both trust and safety in automated vehicles.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• trust calibration
• automation
• automation surprise
• trust in automation foundations
• takeover transitions
• automation complacency bias

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: behavioral performance data, self report data
• Theoretical Contribution: computational model