Application of classification algorithms in the generation of a network intrusion detection model using the KDDCUP99 database

Abstract:

The technological activities must be guaranteed by an acceptable level of security both in the operations carried out by the users and in the data that travel through the infrastructure of the networks, in this research the traffic database is analyzed KDDCUP99 network [Knowledge Discovery in Databases] in whose results were obtained through the conjunction matrices between the algorithms of Neural Networks and K-NN (K Neighbors Neighbors) to determine the best classifier in the training of the model, as a first step the pre-processing was used of the information and the analysis of the users by means of methods of classification of entities and of specific attributes, using techniques and tools of mining of data such as (to remove duplicate values. SimpleKMeans, selection of more significant attributes, elimination of unused characteristics, voracious algorithms and discrete Chi-Square attributes, the division of our database into two random parts continued: one composed of 70% used to train the model and another with the remaining 30% to validate the result; The results allowed us to determine that the most effective model is not the Neural Networks algorithm with 100% of correctly classified instances compared to K-NN with 100%.

Año de publicación:

2020

Keywords:

• K-NN (K Neighbors Neighbors)
• Multilayer Perceptron
• Neural networks

Fuente:

scopus