Conformal prediction aims to determine precise levels of confidence in predictions for new objects using past experience. However, the commonly used exchangeable assumptions between the training data and testing data limit its usage in dealing with contaminated testing sets. In this paper, we develop a series of conformal inference methods, including building predictive sets and inferring outliers for complex and high-dimensional data. We leverage ideas from adversarial flow to transfer the input data to a random vector with known distributions, which enable us to construct a non-conformity score for uncertainty quantification. We can further learn the distribution of input data in each class directly through the learned transformation. Therefore, our approach is applicable and more robust when the test data is contaminated. We evaluate our method, robust flow-based conformal inference, on benchmark datasets. We find that it produces effective prediction sets and accurate outlier detection and is more powerful relative to competing approaches.