Anomaly Detection in Network Traffic Using Deep Learning

With the growing rise of cyber threats and attacks, the use of machine learning and deep learning-based approaches in network anomaly detection has become a necessity. In this study, four deep learning models, including DNN, LSTM, CNN, and Autoencoder, were examined and evaluated on the ISCX VPN dataset. After preprocessing, which included normalization, dimensionality reduction using PCA, handling missing values, etc, the data were fed into the models. The results showed that the LSTM model achieved the best performance among all models, with an F۱-score of ۰.۹۷ and an accuracy of ۰.۹۷, while the Autoencoder model, due to its simpler architecture, delivered weaker performance with an F۱-score of ۰.۶۴ and an accuracy of ۰.۶۹. These findings indicate that memory-based models provide better performance for detecting complex behavioral patterns in network traffic.

کلیدواژه ها:

Deep learning ، DNN ، CNN ، LSTM ، PCA ، Autoencoder