Generation of Anomalous data points

I have a task to generate some anomalous points in a real world dataset with 15 features, and a synthetic dataset of 5 features. I was thinking of using correlation between features, but it'll be a problem to do for 15 features. Thanks!