Anomaly Based Detection for Identifying R2L (Remote to Local) Attacks Using RNN-LSTM in Comparison with ANN for Reducing False Alarm Rate

Abstract

Aim: Detection of the higher false alarm rate using Novel RNN-LSTM is the objective of this work. Materials and Methods: Classification of anomaly based detection is done for identifying remote to local attacks using recurrent neural networks with sample size of (N=52) in which 26 samples are for RNN and 26 samples are for ANN and both the techniques are compared and results are obtained using the G-power value 0.80. Results and Discussion: The proposed work used Novel RNN-LSTM from the NSL-KDD dataset network anomaly detection has accuracy 71% as well as ANN accuracy 66.08%. Significance value becomes 0.006 $(\mathbf{p} < \mathbf{0.05})$. Conclusion: Novel RNN-LSTM gives an accuracy which is better compared with ANN.