Intrusion detection using anomaly detection approach by discovering internet connection episode
In this thesis, the Author conducts an experiment to develop an alternative method to detect intrusion in network environment. This experiment is based on methodology of analyzing connection patterns, which relation of events may represent patterns that describes connection. From this finding of interesting and significant only patterns, an intrusion attempts that appears in incoming connections stream will be detected before it harms the destination host. As fundamental technology to build this system, in this research the Author used a methodology called Frequent Episode Mining, which enable a system to mine pattern from incoming DataStream. To support the methodology, this research also focusing on how to implement a best mechanism for Frequent Episode Rule to mine efficiently from incoming stream, which is usually massive in an instant. The result and findings of this research also discussed in this thesis, which includes numbers of intrusion detected, performance of this system, and accuracy comparison with other intrusion detection systems standards. Benefits of this research are providing an alternative for improving performance of intrusion detection system and broaden knowledge of mining patterns from streaming data.
B00536 | (wh) | Available |
No other version available