《计算机应用》唯一官方网站 ›› 2026, Vol. 46 ›› Issue (4): 1104-1114.DOI: 10.11772/j.issn.1001-9081.2025050548

• 数据科学与技术 • 上一篇    下一篇

基于混淆的自动驾驶仿真测试场景数据保护方法

彭海洋1, 刘天阳1, 计卫星2(), 刘法旺3   

  1. 1.北京理工大学 计算机学院,北京 100081
    2.北京师范大学 人工智能学院,北京 100875
    3.工业和信息化部 装备工业发展中心,北京 100846
  • 收稿日期:2025-05-19 修回日期:2025-07-18 接受日期:2025-07-25 发布日期:2025-08-01 出版日期:2026-04-10
  • 通讯作者: 计卫星
  • 作者简介:彭海洋(2000—),男,重庆人,硕士研究生,主要研究方向:自动驾驶仿真测试
    刘天阳(1995—),女,河北保定人,博士研究生,主要研究方向:程序分析与优化、软件缺陷库构建
    刘法旺(1981—),男,河南信阳人,正高级工程师,博士,主要研究方向:智能网联汽车。
  • 基金资助:
    新一代人工智能国家科技重大专项(2022ZD0116311)

Obfuscation-based protection method for scenario data in autonomous driving simulation testing

Haiyang PENG1, Tianyang LIU1, Weixing JI2(), Fawang LIU3   

  1. 1.School of Computer Science and Technology,Beijing Institute of Technology,Beijing 100081,China
    2.School of Artificial Intelligence,Beijing Normal University,Beijing 100875,China
    3.Equipment Industry Development Center,Ministry of Industry and Information Technology,Beijing 100846,China
  • Received:2025-05-19 Revised:2025-07-18 Accepted:2025-07-25 Online:2025-08-01 Published:2026-04-10
  • Contact: Weixing JI
  • About author:PENG Haiyang, born in 2000, M. S. candidate. His research interests include autonomous driving simulation testing.
    LIU Tianyang, born in 1995, Ph. D. candidate. Her research interests include program analysis and optimization, software defect repository construction.
    LIU Fawang, born in 1981, Ph. D., senior engineer. His research interests include intelligent connected vehicles.
  • Supported by:
    New Generation Artificial Intelligence National Science and Technology Major Project(2022ZD0116311)

摘要:

仿真测试是验证自动驾驶系统安全性和可靠性的重要技术之一。针对该过程中由场景数据明文共享使用导致的数据泄漏问题,提出一种针对场景数据的混淆保护方法。该方法包括数据重编码、命名替换、顺序扰乱、标签重构、触发条件混淆及事件混淆等混淆方法,并按混淆强度划分3个混淆等级,从而在不影响仿真测试结果的情况下提高场景数据的安全性。实验结果表明,混淆后的场景数据在仿真结果上与原始数据基本一致,误差在合理范围内,且随着混淆等级的提高,数据保护程度逐渐增强。一级和二级混淆方法对仿真效率无显著影响,三级混淆方法虽引入了一定的额外计算开销,但仍保持在合理范围内。整体上,三级混淆方法体系能够在保持合理的仿真性能的基础上,有效防止数据泄漏,为自动驾驶仿真测试场景数据保护提供可行的解决方案。

关键词: 自动驾驶, 仿真测试, OpenScenario, 数据安全, 数据混淆

Abstract:

Simulation testing is a critical technology for verifying the safety and reliability of autonomous driving systems. To address the data leakage caused by plaintext shared use of scenario data during this process, an obfuscation protection method for simulation testing scenario data was proposed, along with a corresponding three-tier obfuscation strategy. In this method, a series of obfuscation techniques were covered, including data re-encoding, name replacement, sequence scrambling, label reconstruction, trigger condition obfuscation, and event obfuscation, and it was divided into three obfuscation levels according to obfuscation intensity, thereby enhancing scenario data security significantly without influencing simulation testing results. Experimental results demonstrate that the simulation results for obfuscated scenario data are consistent with simulation results for the original data, and the error of the method is within a reasonable range. As the obfuscation level increases, the degree of data protection also improves progressively. The first and second-level obfuscation methods have no significant impact on simulation efficiency, whereas the third-level method introduces a slight delay in simulation execution time within a reasonable range. Overall, three-level obfuscation method system maintains reasonable simulation performance while preventing data leakage effectively, providing a practical solution for the protection of autonomous driving simulation testing scenario data.

Key words: autonomous driving, simulation testing, OpenScenario, data security, data obfuscation

中图分类号: