Summary: When attacking a distributed protocol, an adaptive adversary may determine its actions (e.g., which parties to corrupt), at any time, based on its entire view of the protocol including the entire communication history. In this paper we are concerned with proactive RSA protocols, i.e., robust distributed RSA protocols that rerandomize key shares at certain intervals to reduce the threat of long-term attacks. Here we design the first proactive RSA system that is secure against an adaptive adversaries. The system achieves “optimal-resilience” and “secure space scalability” (namely $O(1)$ keys per user).