Learning is never secure: Poison learning by Intrusion Detection System based on Self-Organizing Map
Rupam Sharma, Hemanta Kalita, Swagatam Das, Biju Issac · 2016
Machine learning has proven to enhance the detection rate by any Intrusion Detection Engine. However, if learning is tampered than there is a likeliness that the entire detection engine might fail. This paper explores the adversary learning by SOM (Self Organizing Map) trained on NSL-KDD data set. Experimental results demonstrate the deviation of the learning behaviour after being trained with crafted poison packets to mislead the learning.