Anomaly based intrusion detection for a biometric identification system using neural networks

Abstract

This manuscript presents a supervised machine learning approach in the identification of network attacks on a fingerprint biometric system. To reduce the problem of malicious acts on a biometric system, this manuscript proposes an intrusion detection technique that analyses the fingerprint biometric network traffic for evidence of intrusion. The neural network algorithm that imitates the way a human brain works is used in this study to classify normal traffic and learn the correct traffic pattern on a fingerprint biometric system. The aim of the study is to observe the ability of the neural network in the detection of known and unknown attacks without using a vast amount of training data. The results of the neural network model had a classification rate of 98 %, which translates to a false positive rate of 2%.