I am using CSIC 2010 dataset for Network Intrusion Detection. I find it difficult to preprocess the HTTP packet captured. Are there any ways I can easily preprocess data so that good accuracy is achieved?
By preprocessing, if you mean data extraction, then the process is pretty simple.
- Each line of a record is a property-value pair
- You can easily split the property from its value by the colon ':' that separates them
- If the property is real valued, treat the property as a feature
- If the property is nominal, set a feature for each of its possible value and have a boolean value indicate its presence or absence
- The end of a sample record is indicated by a two blank lines (or end-of-file for the last record).
When you already have the data, further preprocessing may involve feature selection and noise reduction.
Multilayer Perceptron (MLP) is used, it is one of the most common and well-known Feedforward Neural Network (FFNN) models.
It permits the data flow to travel one way, from input to output. There is no feedback; it tends to be straight-forward networks that companion inputs with outputs. Any MLP network can be distinguished by a number of performance characteristics, which can be summarized in three points:
- Neural Network Architecture.
- Training Algorithm.
- Transfer Function.
Sigmoid Function is one of the most commonly used transfer functions. The use of the sigmoid function has an advantage in neural networks trained by a backpropagation
learning algorithm, this function responsible for reducing the errors , so it guarantees the security term.
Kindly refer to :
http://manual-snort-org.s3-website-us-east-1.amazonaws.com/node17.html
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Control Theory
- Control Systems