Memory Carving in Embedded Devices: Separate the Wheat from the Chaff.

This paper investigates memory carving techniques for embedded devices. Given that cryptographic material in memory dumps makes carving techniques inefficient, we introduce a methodology to distinguish meaningful information from cryptographic material in small-sized memory dumps. The proposed methodology uses an adaptive boosting technique with statistical tests. Experimented on EMV cards, the methodology recognized 92% of meaningful information and 98% of cryptographic material.