Intrusion Detection System Model Based on Gated Recurrent Unit to Detect Anomaly Traffic

In this study, we proposed the Gated Recurrent Unit method to develop an Intrusion Detection System to detect traffic anomalies. This model was trained in several scenarios to get the best model. Scenario variations were carried out on the learning rate, the activation function and the optimization function. We used the CIC IDS 2017 dataset because it has the latest types of attacks. The dataset was divided into 3 parts, namely training, validation and a test with a ratio of 60%, 20% and 20% respectively. The model was compiled with a binary cross-entropy loss function. Measurement of the results was done using metric accuracy and F1 Score. The experimental result shows that the model with the highest accuracy is obtained from a model that used a learning rate of 0.0001, the LRelu activation function and the Adam optimization function with accuracy and the F1 score reaching 97.7739% and 97.7979% respectively. These results exceed the accuracy of previous studies on the same dataset and the same case classification. On the other hand, it can be concluded also from the study results that the models compiled using the Adam optimized have higher accuracy than the models compiled using Stochastic Gradient Descent. Our studies have not been able to figure out what types of attacks that the model detects. Therefore, we suggest conducting a further study of the multiclass classification case. In addition, further studies are also needed to increase the speed of the model in studying data so that it can be applied in real time to a production environment.