Tester Jobs Requirements (VIDEO GAME): anonymization

Showing posts with label anonymization. Show all posts

Wednesday, September 6, 2017

User data anonymization

Today, in many cases, we have a need of processing (obtaining, recording or holding) user data and that is when privacy issues arise. Data anonymization ensures that even if data is stolen, it cannot be misused. Healthcare and financial industries are the most exposed to threats since such data is at high demand on black markets, but everyone who is keeping any personally identifiable information (PII) and individuals activities should be aware of the risk. Here are a few interesting data breaches that happened:

Poorly anonymized logs reveal NYC cab drivers� detailed whereabouts
Anthem Blue Cross hack: Millions of records breached at health insurer
Database of 191 million U.S. voters exposed on Internet
The entire Turkish citizenship database has allegedly been leaked online

While working on my project I wasnt sure which part of the collected user data are sensitive and need to be anonymized. Ive discovered that there is no general rule for that. Basically, these are the steps:

Identify all occurring information in the dataset.
Define which information can identify an individual by itself or by combining with some other information.
Determine how each type of such information can be anonymized.
Repeat the process to make sure all identifiable information is sanitized and that no piece of information can be used to re-identify individuals by matching it to some other outside sources. Such pieces of information are called quasi-identifiers.

The following attributes are examples of PII:

Personal: national identification number (SSN), date of birth, age, gender, marital status, religion, race, address, zip code, city, state, vehicle registration number, driving license, photographs, iris scan, biometric details, IP address
Financial: credit card number, CVV, account number, balance, credit
Educational: qualifications, university course, school or college studied, year of passing
Contact information: e-mail address, social networking login, telephone number
Medical information: medical history, patient identification number
Employment related: salary

The third step, process of anonymizing user data, is not an easy task. De-identification is based on characteristics of the different fields, so de-identification types of methods can include:

completely removing data,
masking the values (e.g. Donald to Don*l*) or replacing values with random or fake values
encryption (e.g. Donald to *!#$@a),
data for one or more variables are switched with another record so that data set values stay real but are assigned to the wrong people,
creating aliases by applying a one-way hash to the variable so that the hash value cannot be reversed to the original value,
generalization of quasi identifiers (e.g. low population postal code can be aggregated to a larger geographic area - city),
perturbation by data swapping where you swap pairs of data so that data values stay the same but locations and associated records are changed,
k-anonymity technique where you release one record with a particular identifiable information and then you also release several other records who have the same or similar values.

When choosing the right method for the data anonymization we need to be aware that some methods can significantly reduce the utility of the data. For example, masking should be applied only to the fields that wont be used in data analysis like names or email addresses.

There are regulatory guidelines that can help in anonymization process like HIPAA Privacy Rule or EU Data protection and privacy ethical guidelines. However, there remains the potential to re-identify individuals using data if the process is not done promptly. No matter the network and physical security mechanisms, prevention of misuse of sensitive data can be achieved only by ensuring that the right data is getting to the right people in the right format.

download file now