CNN Based Deep Learning Model on Intrusion Detection System to Improve High Accuracy on Big Data

Abstract

Model on Intrusion Detection System to Improve High Accuracy on Big Data on rapid expansion of the digital landscape, the security of networked systems has become a paramount concern. Network intrusions, which involve unauthorized access and malicious activities, pose significant threats to the confidentiality, integrity, and availability of sensitive information. To counter the threats, intrusion detection systems (IDS) play a crucial role in identifying and mitigating such intrusions.

Objectives:  The objective of this research study is to develop and predictive model based on CNN to identify the intrustions during data transamission in real time mode. The researer emphasised the exponential growth of data in digital ecosystems, traditional Intrusion Detection Systems (IDS) often struggle to maintain accuracy and performance when faced with large-scale, high-dimensional datasets. This research proposes a Convolutional Neural Network (CNN)-based deep learning model tailored for Intrusion Detection in Big Data environments.

Methods:   The researchers The model leverages CNN’s ability to automatically learn hierarchical spatial features, enabling efficient detection of complex and subtle patterns associated with cyber threats. By transforming network traffic data into structured forms amenable to CNN processing, the model achieves enhanced feature extraction and classification capabilities. Extensive experiments conducted on benchmark intrusion datasets such as NSL-KDD and CICIDS2017 demonstrate that the proposed CNN-based IDS significantly improves detection accuracy, reduces false positive rates, and scales efficiently with large data volumes.

Results:  The main objective of implementation of CNN algorithm adapted to the context of intrusion detection due to its ability to discover patterns in large datasets.  The researchers found the 99% accuracy level using CNN Basic Performance Model and At 100/100, it takes 63s 631ms/step to lose 0.2421ms per step, and it finds a way 0.850ms per way to acquire 1.05ms.99.9% of the time has elapsed since Epoch started.

Conclusions:  This study underscores the potential of deep learning, particularly CNN architectures, in building robust, scalable, and high-accuracy IDS frameworks suitable for modern Big Data analytics. The researchers found that CNN model on training data and validating it on validation data, it can be interpreted that: Model was trained on 80 epochs and then on 30 epochs, CNN performed exceptionally well on training data and the accuracy was 99%. At this stage, the researchers’ concluded that random forest classifiers gave more accurate results than migration. According to statistics, the confusion matrix score is 80% accuracy, 0.96 "no balance" precision, 0.93 recovery rate, 0.94 F1score and 10000 supporting features.