SAFE-ADVENT

Wireless communication technology is a vital enabler for~5G vehicle-to-everything (5G-V2X) communications, which remain vulnerable to different types of cyberattacks, such as distributed denial of service (DDoS).~\mbox{5G-V2X} networks demand enhanced mechanisms for DDoS detection, which would call for sophisticated artificial intelligence \mbox{(AI)-based} approaches. However, one main concern is data privacy when using \mbox{AI-based} solutions. Secured peer-to-peer (P2P) federated learning (FL) techniques could overcome this challenge by training detection models locally while sharing model parameters for secure aggregation to preserve privacy and bypass the centralized FL's single point of failure. However, securing P2P FL with secure average computation (SAC) or encryption/decryption would scale poorly and incur a high communication cost as the number of FL clients grows. In such a context, we propose \mbox{SAFE-ADVENT}, a novel, secure P2P FL strategy that significantly reduces the communication cost, incorporating transfer learning and client selection within the P2P FL system to detect DDoS attacks. We validate the generalization capability of \mbox{SAFE-ADVENT} using data collected from our~\mbox{5G-V2X} testbed. We demonstrate our method's superior performances, in terms of accuracy, robustness, and cost, across different scenarios with unbalanced and balanced dataset distributions, compared to existing benchmarks.

无线通信技术是5G车联网（5G-V2X）通信的关键支撑技术，但该技术仍易遭受各类网络攻击，例如分布式拒绝服务（DDoS）攻击。5G-V2X网络需要更完善的DDoS检测机制，这就需要采用先进的人工智能（AI）方法。然而，使用基于AI的解决方案时，一个主要问题是数据隐私。安全的点对点（P2P）联邦学习（FL）技术可通过在本地训练检测模型，同时共享模型参数以进行安全聚合，从而克服这一挑战——既能保护隐私，又能避免集中式FL的单点故障问题。然而，若采用安全平均计算（SAC）或加密/解密技术保障P2P FL的安全性，随着FL客户端数量的增加，其扩展性会变差，且通信成本会显著升高。在此背景下，我们提出SAFE-ADVENT——一种新型安全P2P FL策略，该策略通过在P2P FL系统中融入迁移学习和客户端选择机制来检测DDoS攻击，可显著降低通信成本。我们利用从5G-V2X测试平台收集的数据验证了SAFE-ADVENT的泛化能力。与现有基准方法相比，我们的方法在数据集分布平衡和不平衡的不同场景下，均在准确性、鲁棒性和成本方面表现出更优性能。