Considering Context in Procedures of Personal Data Discovery

In a connected and not fully regularized world, the protection of personal data is one of the obligations of government authorities towards their citizens. The GDPR (General Data Protection Regulation) is the European directive instituted to oblige organizations and in particular Internet commercial companies to respect the data concerning their customers and to forge a bond of trust with them. The anonymization of data is one of the methods of complying with GDPR. Protecting sensitive business data, and in particular customer personal data, starts with knowing what sensitive data is and where it is located. Data discovery inspects metadata and real data in different databases to discover personal data and provides comprehensive results listing sensitive fields storing that data. The specification of personal data is a very time-consuming phase spent by business experts in the analysis, especially in the case of databases with large amount of data stored in thousands of fields in the tables. There are data discovery tools that use established procedures to determine personal and sensitive data in applications of organizations and companies. However, these tools cannot always extract the exhaustive list of personal data or incorrectly specify certain data as personal. The error deviation in automatic detection is due to the context in which the data is used. For instance, a field typed as first name will not be considered as storing personal data if all the records in the database for this field always return the values “Zoe” and “Megane” which are certainly the first names of people but in the application context, the values correspond to names given to categories of car models. The procedure for identifying personal data must consider the practices of the business analyst who is familiar with the data included in these applications. These practices correspond to user actions performed in the different contexts of use of data to confirm whether the data is personal. This paper presents how to contextualize procedures used in personal data discovery to improve the anonymization process adopted by a company to protect its corporate applications. The anonymization is reinforced by different contextualization of connected data groups through user practices added to well-established procedures. We use the Contextual Graphs Formalism to improve a procedure of sensitive personal data discovery used in the anonymization process.