Outlier detection using Diffusion models in Tabular data

Question

I have used autoencoders to do outlier detection in tabular data. I want to use diffusion models to do anomaly detection in tabular data. can someone share a resource link or some inputs to do this task.

My idea: Same like VAEs, we calculate the reconstruction error for the train data. If the test data samples does not fall under the reconstruction error distribution as train data, we flag those samples as outliers.