Network Intrusion Detection Using ADASYN and Hybrid Residual Blocks

Computer Engineering

An approach based on adaptive synthesis is presented to address the issues of low accuracy in classification and inadequate small sample feature collection of current intrusion traffic detection models. Enhanced residual network approach that makes use of Inception-Resnet and sampling modules. When applied to unbalanced data sets, this approach can efficiently optimize sampling and enhance the model limited capacity to extract features from a small sample. The unbalanced data training set is first oversampled to enhance the data distribution, then the non-data portion is subsequently separately. To simplify preparation, hot storing is processed and combined with the data portion. Lastly, data training, algorithm performance comparison, and performance evaluation are conducted using the refined residual network model. According to experimental findings, the enhanced residual network model can identify intrusion traffic with an accuracy of between 89.40% and 91.88%. The enhanced residual network model outperforms the traditional deep learning technique for performance, dependability & to detect intrusion.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.