Most User Behavioral Analytics (UBA) applications rely on the distributions and baselines of users and are sensitive to the changes in these patterns. Development and testing of these applications need synthetic data as the availability of the real data is usually scarce. Synthetic data generated must follow these patterns, or else the results can be noisy. Through this work, we present a data generation technique, which could be utilized by UBA applications. The proposed system extracts the patterns of data attributes by considering the dependencies between them. The extracted patterns can be used any number of time to generate data. Additionally, we also generate synthetic users, whose behaviors and distributions are similar to that of real users. Our experiments show that the synthetic data captures the required patterns and relations from the real data. We also show that our data generation process can be scaled linearly to the available processors.