A Feature Selection based Decisive Red Fox Algorithm with Deep Learning for Protecting Cybersecurity Network
Sravanthi Dontu, Santosh Reddy Addula, Piyush Kumar Pareek, Rohith Vallabhaneni, Mohsen Fallah · 2024
There has been a meteoric rise in Internet use recently. Internet use is at an all-time high around the globe. The increasing dependence on the Internet has also raised the risks of damaging assaults. Threats to cyber security arise when individuals or groups engage in harmful activities with the goal to compromise, damage, or otherwise interfere with computer systems. As worries about cybercrime continue to grow, cyber defenses have become an essential tool in the battle against online fraud, forged documents, and other forms of online violence. Intrusion detection systems can detect malicious activity on a network and alert administrators to potential threats. Intrusion Detection Systems (IDS) have adopted a number of methodologies. Some factors contribute to how effective they are. Still, it opens the door to additional research. One of the most pressing issues in cyber defense is the development of an automated system to identify cyberattacks. There has been a recent uptick in the number of studies showing that ML approaches are better than more traditional forms of intrusion detection systems. A feature selection approach that is based on deep learning is utilized in this research effort. Attack detection is the first use of the 5G-NIDD dataset, with Decisive Red Fox (DRF) Optimization utilized to optimally choose the essential structures. In addition to that, it helps the classifier's mistake rate and training speed. Additionally, in order to classify the normal besides attacking data flows using optimum features, the DBRF classification model is utilized. The last step is to classify the data and then test the model's performance using a Convolutional Neural Network (CNN). Specificity, recall, F1-score, accuracy, precision, besides dependability are some of the performance metrics that will be assessed for different types of cyber-attacks as a result of the deployment. Using a variety of evaluation criteria, we compare the findings to those of earlier anomaly detection methods. DRF-CNN perfect got the accuracy as 95.62 besides precision as 98.32 besides recall rate as 94.62 besides then f1-score by means of 94.53 congruently.