A Data Mining Based Approach in IDS Design
Rakesh Kumar Yadav, Mahesh Malviya · 2013
Security is major issue now in these days in different application level as well as in the network level applications and utilities. This paper is based on a new approach based on process mining. In daily use we use various computer based application and interacted through different processes. Some of the process is well known and they provide support for smart works. But some processes are malicious and interrupting different kinds of applications, in this project we are going to introduce the malicious processes classification for using it over IDS development. For that purpose we make efforts for analysing different processes collected from the server to client's machines. I. Introduction In this era of technology their various decisions are taken from the machine, this decision making capability of any machine is developed through machine learning concept. Using the machine learning concept we make such algorithms by which machine make intelligences decisions from the previously provided examples and learning sets. During the study of different kinds of data mining based IDS implementation we found that most of the systems uses the data base and they match data in sequential pattern. Due to matching sequentially the time complexity and memory complexity is effected, thus if we create a data structure and using the parameters we navigate this model than we can save time and memory resources by training these models once. Machine learning works in three main steps first data cleaning or pre-processing in this phase data is read from source and converted into a desired form of data which is easily recognized by the algorithm, in the next step algorithm accepts pre-processed data and build data model on the basis of the input training data set. After that in the third step a number of parameters are provided as input and expect the pattern as a classified way. This process in much similar to the finding a formula for a given series which formula satisfy all the similar kind or same kinds of series, thus we can say the data mining techniques take training data as input and arrange them into a structured form for suitable and easy navigation.