Analysis of partial key recovery attack on multivariate cryptographic transformations using rank systems

Abstract

The Rainbow signature scheme, proposed by Ding and Schmidt in 2005, is one of the oldest and most studied signature schemes in multidimensional cryptography. The Rainbow, based on the unbalanced Oil and Vinegar signature scheme, has the necessary cryptocurrency since 1999 with the right parameters. Interest in multivariate cryptography has increased in the last decade, as it is considered to be quantum-stable. Cryptanalysis of the Rainbow and its predecessors was actively developed in the early 2000s. Attacks from this era include the MinRank attack, the HighRank attack, the Bill-Gilbert attack, the UOV agreement attack, and the Rainbow bandwidth attack. After 2008, cryptanalysis seemed to have stopped, until the Rainbow's participation in the NIST PQC project, which motivated the continuation of cryptanalysis. During the second round of NIST, Bardett and others proposed a new algorithm for solving the MinRank problem. This dramatically increased the effectiveness of MinRank's attack, although not enough to threaten the parameters provided to NIST. A less memory-intensive version of this algorithm was suggested by Baena et al. Perlner and Smith-Tone analyzed the Rainbow bandwidth attack in depth, which showed that the attack was more effective than previously thought. This prompted the Rainbow team to increase slightly the parameters for the third round. During the third round, Bellens introduced a new attack that reduced the Rainbow's security by 220 times for SL 1. The Rainbow team claimed that despite the new attacks, the Rainbow's parameters still met NIST requirement. The purpose of this article is to present two new (partial) key recovery attacks on multivariate cryptographic transformations using rank systems.