What Is Data Perturbation?

Data perturbation is a data mining method that preserves the privacy of electronic health records. It's easy and effective, and it works! Two main data perturbation types are appropriate for EHR data protection: probability distribution and value distortion. The first type, probability distribution, involves using probabilistic methods to create a new dataset from the original one. This allows you to protect sensitive information by changing it so it is still helpful but cannot be used to identify individuals in the dataset. The second type, value distortion, involves changing values within an existing dataset so they cannot be traced back to individuals or groups in the original data set. Value distortion can be done by simply deleting or replacing some values with random numbers; however, this approach may only work well for some data types. The probability distribution approach to perturbation is like taking your car to the mechanic. The mechanic will replace your spark plugs, oil change and transmission fluid. He'll do it with the same brand of parts that came with your car in the first place. The value distortion approach is more like taking your car to a chop shop. You don't know what version of the car you're getting back, but you do know that it's going to be different from what you left with and not just because of the missing parts! As far as data mining goes, both approaches can be effective. It's pretty clear which one we'd prefer. Data pertubation is the next big thing in data protection, and it's here to stay. Why? Because data pertubation is more effective than de-identification/re-identification at protecting health care data. Why? It's harder to break and harder to break, and attacks on this type of data are less likely. For example, what would you do if you were a hacker looking to get into an EHR system and find a specific person's information? You'd try to break into the system and get their information. If the data were all scrambled up, not just scrambled but scrambled in such a way that no one else could understand, then you'd have no idea where any of the pieces went. And you'd have no idea where any of the pieces went… ever!