Novel fault injection attack without artificial trigger

Abstract

Theoretical process of fault injection attacks is defined as a process of recovering a secret key assuming that an attacker can inject faults into a specific targeted operation. Therefore, an artificial triggering is required to execute such an attack. However, when conducting analysis on real devices, artificial triggering needs to rely on a powerful assumption, such as manipulation of internal codes. In this paper, we propose a novel fault injection system using Input/Output (I/O) signals of target devices as a trigger for relaxing an attacker assumption. This system does not require an implementation of artificial triggering as input signals are used as a trigger in transmission of plaintexts for fault injection attacks. As a result, the attacker can perform fault injection attacks concerning the entire encryption process. To decide the fault injection time based on the trigger, the proposed system applies simple power analysis (SPA), employing electromagnetic emission of target devices. Considering that the fault injection time identified by SPA can be relatively vague compared with that obtained using a system based on an artificial triggering, we address this problem by proposing a process to recover the secret key without knowing the byte index of an injected fault.