Analyzing white-box designs for differential computation analysis resistance

Traditionally, cryptographic algorithms are designed to protect data in the black-box attack model, where an attacker can only tinker with the communication channel while the endpoints are trusted. This model fails to capture the full capabilities of the attacker if a cryptographic algorithm runs in an untrusted environment. The attacker may have partial or even full access to the implementation, which enables attacks that are not possible in a black-box context. The white-box attack model considers the case that the attacker controls the environment and has unrestricted access to the implementation. White-box cryptography is concerned with implementing ciphers in a secure manner under this model. All public white-box designs have been broken. Most of the attacks on white-box designs require detailed knowledge of the designs. On the other hand, the Differential Computation Analysis (DCA) attack can quickly retrieve secret key from a number of publicly available white-box implementations without their designs. This attack poses a serious threat to white-box designs in practice. We analyzed a white-box design for the Data Encryption Standard, focusing on its resistance to the DCA attack. The initial design is vulnerable, however, we were able to improve and make it DPA-resistant. We also measured the size and throughput of the implementation. From the measurements, we concluded that the implementation is small in size, and is appropriate to encrypt short messages.