Statistical Disclosure Risk Analysis Service
The ADLS provide a free and confidential Disclosure Risk Analysis Service to enable data holding organisations to confidently release anonymised datasets into the public domain. The service can provide either:
1. An audit of existing statistical disclosure control practices at your organisation.
2. An assessment of the statistical disclosure risk of a dataset(s) to be made available to the public.
What is statistical disclosure?
Statistical disclosure is the association of specific information from an anonymised dataset with a specific population unit (individual, household, business etc). Such a possibility is considered problematic because data has usually been anonymised in order to protect the confidentiality of respondents and therefore if statistical disclosure takes place then that confidentiality has been breached.
The process by which statistical disclosure might take place is often some form of matching where known information is linked to the anonymised datasets through key variables. There are two related processes involved in this: identification – the association of records with specific population units and attribution – the association of particular data values with specific population units. The two processes can happen independently but frequently co-occur.
Disclosure Risk Analysis attempts to measure the risk of a statistical disclosure for a given dataset. There are many different aspects of this. Statistical models are used to estimate how risky a dataset is in broad terms and there are a variety of different techniques for identifying risky records (often called special uniques). Scenario analysis looks at how and why a dataset might be attacked. Then there are methods for analysing the data environment in order to establish what other data might be linked to the anonymised dataset. Collectively these processes provide a disclosure risk profile for an intended data release and pinpoint the actions required to reduce the risk.
When should you use the Statistical Disclosure Risk Analysis Service?
You will need to carry out a disclosure risk analysis if:
1) You are intending to publish a dataset or disseminate it to a third party or you are intending to add new information (e.g. a new variable or new categories on an existing variable) to a dataset that has already published or disseminated.
AND
2) You are legally and/or morally obliged to protect the confidentiality of respondents.
Failure to carry out a sufficient robust disclosure risk analysis could lead to the release of data which is identifiable. Identifiable data could be classed as personal data and therefore such releases could be viewed contravening the data protection act.
For many datasets – for example where a small number of low dimensional datasets are being released – the risk analysis requirements can be quite light and it can be sufficient in such circumstances to carry out a disclosure risk audit, which stops short of full blown statistical analysis of the dataset. With the release of complex microdata, the need for rigorous disclosure risk analyses becomes more acute. If you do not have expertise in house for you will need the services of a disclosure risk analyst.
Further information
For further information on the Statistical Disclosure Risk Analysis Service please contact the ADLS here. You may also like to read our guidance note and report of the disclosure risk work the ADLS conducted for the Department for Communities and Local Government Supporting People datasets.
You can now:
