抽象的な

An Innovative Deep Learning Method for Identifying Anomalies and Preventing Intrusions in Networked Systems

Fnu Ziauddin*

Over the last two to three decades, cyber security has grown significantly in importance due to the remarkable progress and use of computer networks. Large amounts of data are transferred and received across networks, and as these networks have grown, so have the scope and sophistication of assaults. As a result, data is vulnerable to assault while it is transported and stored. To guarantee network security and prevent malware assaults, a strong networked intrusion detection system (IDS) is necessary. In contrast, an IDS is seen as essential to breaches in the availability, privacy, integrity, and confidentiality of data and other resources within the context of network security frameworks. An intrutraffic andon system watches what happens on the network, examines the network traffic, and notifies the system if it detects any odd activity or incursion. Protecting an attacker's network requires anomaly detection as a critical component. Finding threats inside a network by examining its behavior pattern was crucial for many researchers and application frameworks in both IPv4 and IPv6 networks. An effective data mining approach, like machine learning, must be employed to find anomalies. The procedure of gathering data is becoming more significant in this study as it relates to the investigation of the anomalies. For testing purposes, we have used the Knowledge Discovery and Data Mining (KDD) Cup network traffic dataset, which will imitate real-time attack behavior. Panda's data frame has the CSV file loaded, displaying the data in a tabular format. 41 features totaling 10,000 occurrences were used for 4 distinct classes, including "dos", "normal," "probe," and "r2l." For our investigation in this study project, we used two machine learning (ML) approaches and one deep learning methodology. The "Fastai Library" from deep learning has been used by us for intrusion detection categorization. Nonetheless, we have used the Random Forest (RF) and Decision Tree (DT) techniques for analysis in machine learning. Based on accuracy, we have contrasted the deep learning and machine learning models. The Fastai Library has a 92% accuracy rate, Decision Tree has an 82% accuracy rate, and Random Forest has an 84% accuracy rate. Thus, our analysis of accuracy indicates that the deep learning (DL) method may improve the performance of the intrusion detection system (IDS).