GP-GPU를 활용한 다중 시간 단위 네트워크 트래픽 분석

Abstract

When analyzing network traffic data, speed can be of vital importance; especially in regards to network security. The faster an analysis is completed, the more quickly a response can be mounted. This is critical for minimizing the damage caused by attacks on a network, either externally as in the case of a DDoS attack, or internally as in the case of malware installed on the networked machines themselves. In this work, I utilize General Purpose computing on a Graphics Processing Unit (GP-GPU) in order to provide a thorough and detailed analysis in real-time. By taking advantage of the massively multi-core architecture of a GPU, I was able to analyze extremely high packets per second (PPS) counts without the need for expensive and specialized hardware. Furthermore, the data analyzed is processed through multiple time granularities in order to facilitate analysis though a variety of network traffic conditions.