Transfer learning for raw network traffic detection
作者：
Highlights：
• Network intrusion detection models typically learn from expertly crafted features.
• Deep neural network models perform just as well by only examining raw payload bytes.
• Neural networks can feasibly be transferred to constrained edge devices and retrained.
• Transferred detection models reduce computational costs effective at the edge.
• A transferred neural network combined with a classifier outperforms other methods.
摘要
•Network intrusion detection models typically learn from expertly crafted features.•Deep neural network models perform just as well by only examining raw payload bytes.•Neural networks can feasibly be transferred to constrained edge devices and retrained.•Transferred detection models reduce computational costs effective at the edge.•A transferred neural network combined with a classifier outperforms other methods.
论文关键词：Network intrusion detection,Internet of Battlefield Things,Feature engineering,Machine learning,Deep learning,Transfer learning
论文评审过程：Received 9 May 2022, Revised 12 August 2022, Accepted 17 August 2022, Available online 24 August 2022, Version of Record 2 September 2022.
论文官网地址：https://doi.org/10.1016/j.eswa.2022.118641
