Regresar

Anomaly Detection Method in Computer Systems by Means of Machine Learning

Abstract:

The target of this study was to obtain a method of intruders’ detection that offers alternative support for the protection of the computer systems, being necessary to obtain information derived from the LOGS of a host, classified by a Websockets across a computer system and perforations, for his respective evaluation, it was necessary to form the environment to compare the methods relating to the set of information preprocessed, to create the method and his respective putting on the verge of the model, to evaluate it and to implement it as support. The not supervised method was formulated to identify anomalies when the information is not labeled or classified. There were used skills of mining of information of automatic learning developed to implement a system of intruders’ detection with the effective support of statistical specializing hardware and of mining of information: Orange and Python; Libraries for the manipulation and preparation of information: Pycaret, Pandas and Sklearn; Dash for the implementation. The different methods were evaluated establishing the comparative one based on such metric ones of classification as; Counterfoil of confusion, accuracy, precision, and sensibility. The quasi-experimental results of the new method provide few false positives allowing major valuations of detection on having used automatic learning, allowing the effectiveness of helping in the environment of application as support consuming few resources and contributing to the decision making.