Abstract
Based on the Chinese Remainder Theorem (CRT), Quisquater and Couvreur proposed an RSA variant, RSA-CRT, to speed up RSA decryption. Then, Wiener suggested another RSA variant, Rebalanced RSA-CRT, to further accelerate RSA-CRT decryption by shifting decryption cost to encryption cost. However, such an approach makes RSA encryption very timeconsuming because the public exponent e in Rebalanced RSA-CRT is of the same order of magnitude as φ(N). In this paper we study the following problem: does there exist any secure variant of Rebalanced RSA-CRT, whose public exponent e is much shorter than φ(N)? We solve this problem by designing two variants of Rebalanced RSA-CRT, Scheme A and Scheme B. In Scheme A, we focus on designing a variant in which dp and dq are of 160 bits, and e = 2567 + 1. Thus the encryption time is reduced to about 1 2.7 of the time required by the original Rebalanced RSA-CRT. Scheme B is a variant in which dp and dq are of 198 bits, and e = 2511 + 1. Thus its encryption is about 3 times faster than that of Rebalanced RSA-CRT, but the decryption is a little slower than that of Rebalanced RSA-CRT.