摘要 Abstract 

LDoS可以讓攻擊封包隱藏在正常的NB-IoT資料流中來躲避檢測。已經有許多基於CNN的檢測方法被提出。但這些方法往往因訓練資料不多樣化使得模型容易陷入過擬合，本演講將提出更好的解決方案以實現更好的LDoS檢測品質。

LDoS can hide attack packets in normal NB-IoT data flow to avoid detection. Many CNN-based detection methods have been proposed. However, these methods often make the model prone to overfitting due to the lack of diversity of training data. This speech will propose a better solution to achieve better LDoS detection quality.