The data protection impact assessment is used to verify the necessity, proportionality and risks of data processing. Our work is based on the data processed by the technical support of a Wireless Service Provider. The team of WISP tech support uses a machine learning system to predict failures. The goal of our the experiments was to evaluate the DPIA with personal data and without personal data. In fact, in a first scenario, the experiments were conducted using a machine learning application powered by non-anonymous personal data. Instead in the second scenario, the data was anonymized before feeding the machine learning system. In this article we evaluate how much the Data Protection Impact Assessment changes when moving from a scenario with raw data to a scenario with anonymized data.
Evaluating the Impact of Data Anonymization in a Machine Learning Application
Mastroianni M.;
2022-01-01
Abstract
The data protection impact assessment is used to verify the necessity, proportionality and risks of data processing. Our work is based on the data processed by the technical support of a Wireless Service Provider. The team of WISP tech support uses a machine learning system to predict failures. The goal of our the experiments was to evaluate the DPIA with personal data and without personal data. In fact, in a first scenario, the experiments were conducted using a machine learning application powered by non-anonymous personal data. Instead in the second scenario, the data was anonymized before feeding the machine learning system. In this article we evaluate how much the Data Protection Impact Assessment changes when moving from a scenario with raw data to a scenario with anonymized data.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
