[pavel_lishin]

I have a feeling that datasets that are largely defined by human psychology - the list of iPhone passwords is a good example - are less likely to adhere to Benford's law than "naturally" generated datasets.