Real-Time Intrusion Detection based on Network Monitoring and Machine Learning
Subhadeep Chakraborty · International Journal of Computer Applications · 2023
The threat of cyber-attacks continues to grow in significance as our world becomes increasingly reliant on digital communication and data storage.Network intrusion detection aims to prevent unauthorized access, protect sensitive information and preserve digital systems' integrity.In this indepth research study, the network-based intrusion detection methods for identifying suspicious and malicious network activities have been examined.By utilizing network monitoring techniques to capture and collect data, along with machine learning algorithms for classifying intrusive packets, network intrusion detection has been performed.Initially, Wireshark has been employed for network monitoring, successfully capturing a total of 28,889 packets and storing them in PCAP format.This PCAP data was then used for preliminary analysis to investigate the properties of the packets.Following this, the PCAP data was converted into a Comma Separated File format, facilitating the application of machine learning algorithms for further examination.During data preparation, Principal Component Analysis has been employed to detect and eliminate outliers, along with the Chi-Squared method for feature selection.The research involved testing five machine learning models on the processed data, to achieve the most effective model for intrusion detection.Ultimately, the Random Forest algorithm has excelled, boasting an outstanding accuracy rate of 99.43%.This comprehensive investigation highlights the immense potential of network-based intrusion detection methods for uncovering and addressing cyber threats on both a large and small scale.