RSA with Balanced Short Exponents and Its Application to Entity Authentication

Abstract

In typical RSA, it is impossible to create a key pair (e,d) such that both are simultaneously much shorter than φ (N). This is because if d is selected first, then e will be of the same order of magnitude as φ (N), and vice versa. At Asiacrypt'99, Sun et al. designed three variants of RSA using prime factors p and q of unbalanced size. The first RSA variant is an attempt to make the private exponent d short below N0.25 and N0.292 which are the lower bounds of d for a secure RSA as argued first by Wiener and then by Boneh and Durfee. The second RSA variant is constructed in such a way that both d and e have the same bit-length $\frac{1}{2}\log _{2}N+56$. The third RSA variant is constructed by such a method that allows a trade-off between the lengths of d and e. Unfortunately, at Asiacrypt'2000, Durfee and Nguyen broke the illustrated instances of the first RSA variant and the third RSA variant by solving small roots to trivariate modular polynomial equations. Moreover, they showed that the instances generated by these three RSA variants with unbalanced p and q in fact become more insecure than those instances, having the same sizes of exponents as the former, in RSA with balanced p and q. In this paper, we focus on designing a new RSA variant with balanced d and e, and balanced p and q in order to make such an RSA variant more secure. Moreover, we also extend this variant to another RSA variant in which allows a trade-off between the lengths of d and e. Based on our RSA variants, an application to entity authentication for defending the stolen-secret attack is presented.