Towards Low-Perturbation Anonymity Preserving Pattern Discovery