Re: keystroke authentication




pp wrote:
Hi,

Iam very much new to this tool box.My aim is to create and train a
neural network in order to recognise a person by his typing
pattern.ie,i have a 10*13 matrix consisting of 13 input values
including the duration and latency of keystrokes taken 10 times.I
need to train the network with this and then recognise a new pattern
,again consisting of 13 inputs.What are the steps that i should
follow related to creating the network,training and testing it?
Thanks,pp.

There is not much you can do with 13 variables and only 10 cases.
(Search the archives using greg-heath Neq Nw).
You have 1 case per person? Eliminate useless variables and get
more cases per person. A nearest neighbor algorithm can be
implemented using PNN.

Hope this helps.

Greg

.