Hands-On Deep Learning for IoT

Model performance (use case two)

We have tested the autoencoder model on the three datasets for three different layer's IDS. The following screenshot presents the evaluation result snapshot for the application layer's IDS:

As we can see from the screenshot, training accuracy and validation and test accuracy are well above 90% when we used the first 12 most important features for the training. The performance could be different if we used a different feature set.

The following graph presents epoch-wise training accuracy of the preceding model on the application layer IDS dataset:

We obtained some interesting evaluation results for the network ...