An Internet of Things Intrusion Detection Method Based on CNN-FDC
Yahong Ma, Yang Qin, Yujie Gao · 2021
In order to overcome the problem of weak function and low accuracy when the traditional intrusion detection system is faced with complex and high-dimensional network data characteristics. This paper proposes a CNN-FDC (CNN + focus loss + Dropout-Connect) detection method based on convolutional neural networks. After converting the KDD-CUP99 data set to grayscale images, the Dropout-Connect method is used to enhance the model’s ability to eliminate overfitting, and then Focal Loss is used as a loss function to solve the problem of sample imbalance in the data set. From what have been discussed above, it can be concluded that, compared with the traditional LeNet-5 and KNN machine learning algorithms, the model adding the confusion matrix judgment performs better and relatively high accuracy rate.