Single-Trace Attack on NTRU Decryption with Machine Learning and Template Profiling

NTRU is a post-quantum public key cryptosystem based on lattices and was a finalist in the 3rd round of the post-quantum standardization process organized by the National Institute of Standards and Technology (NIST). We present a new single-trace attack against the NTRU decryption algorithm, meaning we can obtain the private key only from one power trace, assuming the profiling phase was conducted before the attack. The attack was performed on the submission implementation from the standardization process. We used several machine-learning methods and Template attacks for the profiling and testing phase. Both methods achieve high accuracy reaching almost 100% tested on more than 12 000 traces. We also provide a comparison of their success score with regard to a different number of points of interest.