Reference no: EM132608908
IMAT5235 Artificial Neural Networks - De Montfort University
Assignment - A neural network model for detecting intrusions or attacks on a computer network
Learning outcome 1: Experience creating an ANN to solve a intrusion attack problems
Learning outcome 2: Experience pre-processing large data sets
Learning outcome 3: Experience using Matlab
Learning outcome 4: Experience using different ANN algorithms implemented in Matlab using the ANN tool box.
Tasks to be undertaken:
The aim of this project is to develop and test a neural network model to be able to detect network intrusions. The neural network model should be capable of distinguishing between ``bad'' connections, called intrusions or attacks, and ``good'' normal connections.
In order to develop this neural network model, you will use the KDD Cup 1999 Data set which was used for The Third International Knowledge Discovery and Data Mining Tools Competition.
The database contains several intrusions simulated in a military network environment.
In order to simplify things, the construction of the neural network model will be based entirely (training and testing) on kddcup.data_10_percent.gz data subset which contains a dataset sample of only 10 % of the entire dataset.
The web-link above contains all the dataset descriptions and details in terms of inputs an outputs.
Task
You have a number of tasks:
• Grab the data set kddcup.data_10_percent.gz from the web site to produce a data set for training and testing your neural network.
• Split the available data sets for training and testing the neural network. It is up to you to decide how many of the inputs are needed as well as how much data is needed to construct the neural network (training and testing data sets). You might wish to use cross validation.
• Train and test a neural network that can identify good and bad connections (intrusion or attack).
We will be looking for:
• Sensible manipulation of the data.
• Sensible choice of training and test data.
• Appropriate choice of network topology.
• Careful thought about the various design parameters for a neural network.
Attachment:- Artificial Neural Networks.rar