标题： 基于联邦学习的路况分类保障 显示英文标题

标题： Safeguarding Federated Learning-based Road Condition Classification

摘要： 联邦学习（FL）已成为一种有前景的解决方案，用于隐私保护的自动驾驶，特别是基于摄像头的道路状况分类（RCC）系统，利用车载的分布式传感、计算和通信资源，而无需共享敏感图像数据。 然而，FL-RCC框架的协作性质引入了新的漏洞：目标标签翻转攻击（TLFAs），其中恶意客户端（车辆）故意更改其训练数据标签以破坏学习模型的推理性能。 此类攻击可能导致车辆将滑腻、危险的道路状况错误分类为良好，并超过推荐速度。 然而，针对基于FL的RCC系统的TLFAs研究仍较为缺乏。 我们通过三项贡献来解决这一挑战：1）我们揭示了现有FL-RCC系统对TLFAs的脆弱性；2）我们引入了一种基于标签距离的度量方法，以精确量化TLFAs带来的安全风险；3）我们提出了FLARE，这是一种防御机制，利用输出层的逐神经元分析来缓解TLFA的影响。 在三个RCC任务、四个评估指标、六个基线和三个深度学习模型上的广泛实验表明，TLFAs对FL-RCC系统的严重性以及FLARE在减轻攻击影响方面的有效性。 显示英文摘要

摘要： Federated Learning (FL) has emerged as a promising solution for privacy-preserving autonomous driving, specifically camera-based Road Condition Classification (RCC) systems, harnessing distributed sensing, computing, and communication resources on board vehicles without sharing sensitive image data. However, the collaborative nature of FL-RCC frameworks introduces new vulnerabilities: Targeted Label Flipping Attacks (TLFAs), in which malicious clients (vehicles) deliberately alter their training data labels to compromise the learned model inference performance. Such attacks can, e.g., cause a vehicle to mis-classify slippery, dangerous road conditions as pristine and exceed recommended speed. However, TLFAs for FL-based RCC systems are largely missing. We address this challenge with a threefold contribution: 1) we disclose the vulnerability of existing FL-RCC systems to TLFAs; 2) we introduce a novel label-distance-based metric to precisely quantify the safety risks posed by TLFAs; and 3) we propose FLARE, a defensive mechanism leveraging neuron-wise analysis of the output layer to mitigate TLFA effects. Extensive experiments across three RCC tasks, four evaluation metrics, six baselines, and three deep learning models demonstrate both the severity of TLFAs on FL-RCC systems and the effectiveness of FLARE in mitigating the attack impact.