This paper is about using Deep Neural networks to identify malicious activities in any server. We have taken a sample web Application case study. We have researched & implemented a methodology by which our system learns customer behavior real time. We do this by tapping data from server logs / event logs and classifying them as malicious behavior or not. We are tapping server data using GROK parser and forwarding the data using LogStash forwarder. The data is organized using advanced elastic search methods. This data is then fed into a long short term memory ( LSTM ) network . Using the algorithm that we have built on top of this AI library – we dynamically detect & report malicious user/customer activities on the server.