Scholarlink

<p>The integration of 5G networking into Software-Defined Vehicles (SDVs) and Vehicular Ad-hoc Networks (VANETs) enables low-latency and high-reliability vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications, supporting safety–critical and data-intensive vehicular services. However, the increased connectivity and architectural flexibility also expose vehicular networks to Denial-of-Service (DoS) attacks that threaten communication availability and transportation safety. This paper presents TransBoostDoS, a hybrid learning-based detection framework for identifying DoS attacks in 5G-enabled VANETs. The proposed approach employs a Transformer-based model to extract temporal features from vehicular message sequences, capturing dynamic behavioral patterns, which are subsequently classified using XGBoost. Experimental evaluations conducted on a balanced VeReMi dataset using five-fold cross-validation across six classes, including Normal, DoS, DoS disruptive, DoS disruptive Sybil, DoS random, and DoS random Sybil. The proposed framework achieved 99.67% overall accuracy, with macro-averaged precision, recall/TPR, and F1-score of approximately 99.66%, 99.66%, and 99.67%, respectively, while maintaining a low false-alarm rate for normal traffic. These results demonstrate the effectiveness of TransBoost-DoS under the balanced VeReMi evaluation setting, although further validation using additional datasets and real-world vehicular traces is required to assess broader generalizability.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A transformer-XGBoost hybrid framework for multi-type DoS detection in 5G-V2X vehicular networks

  • 期刊论文

  • 发表日期：

    2026年7月3日

查看全文
  • Shiang-Jiun Chen,
  • Wen-Bin Hsieh,
  • Hsiang-Yu Tseng

摘要

The integration of 5G networking into Software-Defined Vehicles (SDVs) and Vehicular Ad-hoc Networks (VANETs) enables low-latency and high-reliability vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications, supporting safety–critical and data-intensive vehicular services. However, the increased connectivity and architectural flexibility also expose vehicular networks to Denial-of-Service (DoS) attacks that threaten communication availability and transportation safety. This paper presents TransBoostDoS, a hybrid learning-based detection framework for identifying DoS attacks in 5G-enabled VANETs. The proposed approach employs a Transformer-based model to extract temporal features from vehicular message sequences, capturing dynamic behavioral patterns, which are subsequently classified using XGBoost. Experimental evaluations conducted on a balanced VeReMi dataset using five-fold cross-validation across six classes, including Normal, DoS, DoS disruptive, DoS disruptive Sybil, DoS random, and DoS random Sybil. The proposed framework achieved 99.67% overall accuracy, with macro-averaged precision, recall/TPR, and F1-score of approximately 99.66%, 99.66%, and 99.67%, respectively, while maintaining a low false-alarm rate for normal traffic. These results demonstrate the effectiveness of TransBoost-DoS under the balanced VeReMi evaluation setting, although further validation using additional datasets and real-world vehicular traces is required to assess broader generalizability.