Big Data Analysis for Intrusion Detection System in Network with Attention based Hybrid Deep Learning Model

Intrusion Detection System (IDS) in network plays crucial role to detect the cyberattacks malicious activities like data bleaching and unauthorised access of secured data. To overcome these drawbacks, detecting the network intrusions at early stage is necessary which is proposed in this research. Using big data analysis, an Attention based Hybrid Deep Learning model is proposed for Intrusion Detection System (IDS) in network. The UNSW-NB15 dataset is considered with intrusion sample entries and pre-processed using min-max normalization as well as feature vectorization employing one-hot encoding technique. The pre-processed data is then fed to Convolutional Neural Networks (CNN) to detect spatial features followed by the Long Short-Term Memory (LSTM) network for the temporal dependencies in the sequential information. An attention mechanism is also incorporated to enhance the classification process to focus on crucial features thus enabling accurate detection. Experimental results showed that the proposed Attention based CNN-LSTM (Att-CNN-LSTM) achieved better detection accuracy with 99.76%, precision 97.62%, and F1-score 96.49% when compared with the traditional DL based IDS such as Fully Streaming Big Data Framework (FSBDL) and Deep Convolutional Nural Network with Bidirectional LSTM (DCNN-BiLSTM).