We propose a method to study the secrecy constraints in key generation problems where side information might be present at untrusted users. Our method is inspired by a recent work of Hayashi and Tan who used the Rényi divergence as the secrecy measure to study the output statistics of applying hash functions to a random sequence. By generalizing the achievability result of Hayashi and Tan to the multi-terminal case, we obtain the output statistics of applying hash functions to multiple random sequences, which turn out to be an important tool in the achievability proof of strong secrecy capacity regions of key generation problems with side information at untrusted users. To illustrate the power of our method, we derive the capacity region of the multiple private key generation problem with an untrusted helper for continuous memoryless sources under Markov conditions. The converse proof of our result follows by generalizing a result of Nitinawarat and Narayan to the case with side information at untrusted users.