Dubrov and Ishai (STOC 2006) revealed, by generalizing the notion of pseudorandom generators (PRGs), that under a computational assumption, randomness in a protocol can be replaced with pseudorandomness in an indistinguishable way for an adversary even if his algorithm has unbounded complexity. However, their argument was applied only to some special protocols. In this article, we first show that their argument is not effective for a wide class of more general protocols. Then we propose a novel evaluation technique for such indistinguishability that is based on usual PRGs and is effective for those more general protocols. Examples of such protocols include parallel computation over honest-but-curious modules, secret sharing, broadcast encryption, traitor tracing, and collusion-secure codes.