Advanced

Change of Purpose - The effects of the Purpose Limitation Principle in the General Data Protection Regulation on Big Data Profiling

Westermann, Hannes LU (2018) JURM02 20181
Department of Law
Faculty of Law
Abstract
Over the past few years, many companies have started to adopt Big Data technologies. Big Data is a method and technology that allows the collection and analysis of huge amounts of all kinds of data, mainly in digital form. Big Data can be used, for example, to create profiles of online shopping users to target ads. I call this Big Data Profiling. Facebook and Google, for example, are able to estimate attributes, such as gender, age and interests, from data provided by their users. This can be worrisome for many users who feel that their privacy is infringed when the Big Data Profiling companies, for example, are able to send advertisements to the users that are scarily relevant to them.

Big Data Profiling relies on a vast amount of... (More)
Over the past few years, many companies have started to adopt Big Data technologies. Big Data is a method and technology that allows the collection and analysis of huge amounts of all kinds of data, mainly in digital form. Big Data can be used, for example, to create profiles of online shopping users to target ads. I call this Big Data Profiling. Facebook and Google, for example, are able to estimate attributes, such as gender, age and interests, from data provided by their users. This can be worrisome for many users who feel that their privacy is infringed when the Big Data Profiling companies, for example, are able to send advertisements to the users that are scarily relevant to them.

Big Data Profiling relies on a vast amount of collected data. Often, at the time of collection, it is not clear how exactly this data will be used and analyzed. The new possibilities with Big Data Profiling have led to companies collecting as much data as possible, and then later figuring out how to extract value from this data. This model can be described as “collect-before select”, since the data is first collected, and then “mined” for correlations that can be used to profile users.

In this thesis I analyze whether this form of collection and usage of Personal Data is legal under the General Data Protection Regulation (GDPR), which enters into force in the European Union on 25 May 2018. While many of the provisions of the GDPR already existed in the Data Protection Directive (DPD) since 1995, they have been reinforced and extended in the GDPR.

One of the main principles of the GDPR is that of Purpose Limitation. While the principle already exists under the DPD in a very similar fashion, it is likely that it will be enforced more under the GDPR, since the GDPR is directly applicable in member states instead of having to be implemented. The enforcement mechanisms, such as sanctions, have also been significantly strengthened.

The Purpose Limitation Principle requires the data controller (such as companies processing Personal Data, like Facebook and Google) to have a specified purpose for and during the collection of Personal Data. Further, the Personal Data cannot be processed beyond this purpose after it has been collected. This seems to run contrary to Big Data Profiling, which regularly looks for purposes only after the Personal Data has been collected. However, I have identified three potential ways the “collect before select” model could still be possible under the GDPR.

The first possibility is the anonymization of Personal Data. If data can be efficiently anonymized, it will fall outside of the scope of the GDPR because it will not contain Personal Data after the anonymization. The controller is then free to analyze the data for any purpose, including creating models that could be used to profile other users. However, I found that Big Data methods can often reidentify Personal Data that has been previously anonymized. In such cases even purportedly anonymized data may still fall under the scope of the GDPR. If on the other hand enough Personal Data is removed to make reidentification impossible, the value of the data for large parts of the business world is likely destroyed.

The second possibility is collecting Personal Data for a specified purpose that is defined so widely that it covers all potential future use cases. If a controller can collect Personal Data for a vague purpose, such as “marketing”, the controller will have a lot of flexibility in using the data while still being covered by the initial purpose. I found that the GDPR requires data controllers (such as companies) to have a purpose for the data collection that is specific enough so that the data subject is able to determine exactly which kinds of processing the controller will undertake. Having a non-existent or too vague purpose is not sufficient under the GDPR. Companies that collect data with no, or an only vaguely defined, purpose and then try to find a specific purpose for the collected data later will therefore have to stop this practice.

The third possibility can be used if the controller wants to re-use Personal Data for further purposes, after the controller has collected the Personal Data initially in compliance with the GDPR for a specified purpose. In this case, the GDPR offers certain possibilities of further processing this data outside of the initial purpose. The GDPR allows this for example if the data subject has given consent to the new purpose. However, I found that Big Data Profiling companies often come up with new purposes later by “letting the data speak”, which means by analyzing the data itself to find new purposes. Before performing an analysis, often the company might not even know how the processing will be done later. In that case, it is impossible to request the data subject’s specific consent, which is required under the GDPR. Even without the data subject’s consent, there are however other possibilities of further processing data under the GDPR, such as determining whether the new processing is compatible with the initial purpose. My thesis examines some of those possibilities for a change of purpose under Big Data Profiling.

My conclusion is that the GDPR likely means a drastic impact and limitation on Big Data Profiling as we know it. Personal Data cannot be collected without a purpose or with a vague purpose. Even Personal Data that was collected for a specific purpose cannot be re-used for another purpose except for in very few circumstances. Time will tell how the courts interpret the GDPR and decide different situations, how the companies will adapt to them and if the legislator will react to this reality. (Less)
Please use this url to cite or link to this publication:
author
Westermann, Hannes LU
supervisor
organization
course
JURM02 20181
year
type
H3 - Professional qualifications (4 Years - )
subject
keywords
EU law, IT law, Artificial Intelligence, Big Data, Profiling, Big Data Profiling, Google, Facebook, GDPR, General Data Protection Regulation, Purpose Limitation, Further Use, Purpose Specification, Personal Data, Compatible Use, Machine Learning, Deep Learning
language
English
id
8941820
date added to LUP
2018-06-10 15:25:36
date last changed
2018-06-10 15:25:36
@misc{8941820,
  abstract     = {Over the past few years, many companies have started to adopt Big Data technologies. Big Data is a method and technology that allows the collection and analysis of huge amounts of all kinds of data, mainly in digital form. Big Data can be used, for example, to create profiles of online shopping users to target ads. I call this Big Data Profiling. Facebook and Google, for example, are able to estimate attributes, such as gender, age and interests, from data provided by their users. This can be worrisome for many users who feel that their privacy is infringed when the Big Data Profiling companies, for example, are able to send advertisements to the users that are scarily relevant to them.

Big Data Profiling relies on a vast amount of collected data. Often, at the time of collection, it is not clear how exactly this data will be used and analyzed. The new possibilities with Big Data Profiling have led to companies collecting as much data as possible, and then later figuring out how to extract value from this data. This model can be described as “collect-before select”, since the data is first collected, and then “mined” for correlations that can be used to profile users.

In this thesis I analyze whether this form of collection and usage of Personal Data is legal under the General Data Protection Regulation (GDPR), which enters into force in the European Union on 25 May 2018. While many of the provisions of the GDPR already existed in the Data Protection Directive (DPD) since 1995, they have been reinforced and extended in the GDPR. 

One of the main principles of the GDPR is that of Purpose Limitation. While the principle already exists under the DPD in a very similar fashion, it is likely that it will be enforced more under the GDPR, since the GDPR is directly applicable in member states instead of having to be implemented. The enforcement mechanisms, such as sanctions, have also been significantly strengthened. 

The Purpose Limitation Principle requires the data controller (such as companies processing Personal Data, like Facebook and Google) to have a specified purpose for and during the collection of Personal Data. Further, the Personal Data cannot be processed beyond this purpose after it has been collected. This seems to run contrary to Big Data Profiling, which regularly looks for purposes only after the Personal Data has been collected. However, I have identified three potential ways the “collect before select” model could still be possible under the GDPR.

The first possibility is the anonymization of Personal Data. If data can be efficiently anonymized, it will fall outside of the scope of the GDPR because it will not contain Personal Data after the anonymization. The controller is then free to analyze the data for any purpose, including creating models that could be used to profile other users. However, I found that Big Data methods can often reidentify Personal Data that has been previously anonymized. In such cases even purportedly anonymized data may still fall under the scope of the GDPR. If on the other hand enough Personal Data is removed to make reidentification impossible, the value of the data for large parts of the business world is likely destroyed.

The second possibility is collecting Personal Data for a specified purpose that is defined so widely that it covers all potential future use cases. If a controller can collect Personal Data for a vague purpose, such as “marketing”, the controller will have a lot of flexibility in using the data while still being covered by the initial purpose. I found that the GDPR requires data controllers (such as companies) to have a purpose for the data collection that is specific enough so that the data subject is able to determine exactly which kinds of processing the controller will undertake. Having a non-existent or too vague purpose is not sufficient under the GDPR. Companies that collect data with no, or an only vaguely defined, purpose and then try to find a specific purpose for the collected data later will therefore have to stop this practice.

The third possibility can be used if the controller wants to re-use Personal Data for further purposes, after the controller has collected the Personal Data initially in compliance with the GDPR for a specified purpose. In this case, the GDPR offers certain possibilities of further processing this data outside of the initial purpose. The GDPR allows this for example if the data subject has given consent to the new purpose. However, I found that Big Data Profiling companies often come up with new purposes later by “letting the data speak”, which means by analyzing the data itself to find new purposes. Before performing an analysis, often the company might not even know how the processing will be done later. In that case, it is impossible to request the data subject’s specific consent, which is required under the GDPR. Even without the data subject’s consent, there are however other possibilities of further processing data under the GDPR, such as determining whether the new processing is compatible with the initial purpose. My thesis examines some of those possibilities for a change of purpose under Big Data Profiling.

My conclusion is that the GDPR likely means a drastic impact and limitation on Big Data Profiling as we know it. Personal Data cannot be collected without a purpose or with a vague purpose. Even Personal Data that was collected for a specific purpose cannot be re-used for another purpose except for in very few circumstances. Time will tell how the courts interpret the GDPR and decide different situations, how the companies will adapt to them and if the legislator will react to this reality.},
  author       = {Westermann, Hannes},
  keyword      = {EU law,IT law,Artificial Intelligence,Big Data,Profiling,Big Data Profiling,Google,Facebook,GDPR,General Data Protection Regulation,Purpose Limitation,Further Use,Purpose Specification,Personal Data,Compatible Use,Machine Learning,Deep Learning},
  language     = {eng},
  note         = {Student Paper},
  title        = {Change of Purpose - The effects of the Purpose Limitation Principle in the General Data Protection Regulation on Big Data Profiling},
  year         = {2018},
}