An improved density peaks clustering algorithm based on grid screening and mutual neighborhood degree for network...

Area of Science:

• Computer Science
• Cybersecurity
• Data Mining

Background:

• Network anomaly detection faces challenges due to rapid technological development and increasing abnormal traffic.
• Existing supervised methods fail to detect unknown attacks, while unsupervised methods exhibit low accuracy.
• There is a need for advanced methods to improve the accuracy and efficiency of network anomaly detection.

Purpose of the Study:

• To propose a novel clustering-based network anomaly detection model.
• To introduce a new density peaks clustering algorithm, DPC-GS-MND, for enhanced anomaly detection.
• To address the limitations of existing supervised and unsupervised methods in detecting unknown network threats.

Main Methods:

• Developed a novel density peaks clustering algorithm (DPC-GS-MND) incorporating grid screening and mutual neighborhood degree.
• Utilized grid screening to reduce computational complexity and improve efficiency.
• Employed mutual neighborhood degree to enhance clustering accuracy and defined a cluster center decision value for automatic center selection.

Main Results:

• Experimental results on KDDCup99 and CIC-IDS-2017 datasets demonstrate superior performance of DPC-GS-MND.
• The proposed algorithm achieves higher accuracy in detecting network anomaly traffic compared to existing methods.
• DPC-GS-MND shows improved efficiency in identifying anomalies.

Conclusions:

• The DPC-GS-MND algorithm offers a significant advancement in network anomaly detection.
• It effectively detects unknown network attacks with enhanced accuracy and efficiency.
• The algorithm shows strong potential for application in network anomaly detection systems within complex network environments.