Evaluation of Detection System of Fault Attacks based on Neural Network into a Java Virtual Machine

Abstract

The Java Card technology provides a secure environment for developing smart card application based on Java while also respecting some constraints such as the limited memory and processing card. In addition to the security and cryptography APIs offered by the Java Card technology, the smart card is protected against some threats. But, the fault attacks based on the variation of the physical parameters are able to disrupt its operation. In order to enhance the smart card security, we thought to add an intelligent component able to distinguish between the smooth functioning and the attack. This component is a Neural Network that we developed in C language and integrated in open source Virtual Machine (Avian) in order to simulate the attack effect and the network behavior. In this context, the detection rate of the attacks is 96% with no false positive.