报告详情
  1. 活动首页
  2. 报告清单
  3. 报告详情
Packet Transformer-Hierarchical CNN Fusion for Payload-Independent Multi-Class Intrusion Detection: Extensive NS-3 Evaluation of Fixed and Adaptive Fusion
编号：166 访问权限：仅限参会人 更新：2026-11-10 17:01:00 浏览：7次 Online

报告开始：2026年11月20日 14:30（Asia/Kolkata）

报告时间：15min

所在会场：[S4] Artificial Intelligence and Intelligent Computing [S4-3] Artificial Intelligence and Intelligent Computing

暂无文件

摘要
Encrypted and payload-restricted networks motivate intrusion-detection methods based on observable flow behavior rather than application content. This paper presents an extensive evaluation of a dual-branch architecture combining a Packet Transformer Encoder (PTE) for global feature interactions and a Hierarchical 1D Convolutional Neural Network (HCNN) for local multi-scale structure. Five configurations are compared on 7,200 balanced NS-3-generated flows spanning 15 benign/attack-like traffic profiles and 77 numerical descriptors. The leakage-controlled split contains 4,608 training, 1,152 validation, and 1,440 held-out test flows. Fixed Fusion produced the strongest test result with 98.472% accuracy, 98.483% precision, 98.472% recall, and 98.471% weighted/macro F1. HCNN and PTE reached 97.419% and 97.372% weighted F1, whereas AHFF-PTE reached 96.512%. The adaptive gate remained balanced with mean PTE weight 0.495, showing that lower generalization was not caused by branch collapse. Detailed class analysis shows Infiltration as the hardest class (85.06% F1), while four classes achieved 100% recall. Network measurements maintained mean PDR above 97% for complete 10-, 20-, and 30-node triplicate runs. The results demonstrate complementary local/global representations while showing that fixed fusion can outperform higher-capacity adaptive fusion under the available simulation regime.
 
关键词
encrypted traffic, intrusion detection, Packet Transformer Encoder, hierarchical CNN, adaptive feature fusion, fixed fusion, NS-3.
报告人
HARINATH K R
Research Scholar Jawaharlal Nehru Technological University Anatapuramu

稿件作者
HARINATH K R Jawaharlal Nehru Technological University Anatapuramu
G KISHOR KUMAR Nandyal；Rajeev Gandhi Memorial College of Engineering & Technology
发表评论
验证码 看不清楚，更换一张
全部评论
重要日期

  • 会议日期

    11月20日

    2026

    至

    11月22日

    2026

  • 10月31日 2026

    初稿截稿日期

  • 11月07日 2026

    提前注册日期

  • 11月22日 2026

    注册截止日期

主办单位
United Societies of Science
承办单位
Pondicherry University
协办单位
IEEE Section
IEEE Madras Section
联系方式
登录查看完整联系方式
移动端
在手机上打开
小程序
打开微信小程序
客服
扫码或点此咨询