Evaluating the Impact of Data Anonymization in a Machine Learning Application
Venue & metadata
- Journal/Proceedings: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
- Volume: 13380 LNCS
- Pages: 389 – 400
- Note: Cited by: 1
- Author keywords: Data disappearance; DPIA; GDPR; Illegitimate access to data; Privacy; Risks; Unwanted modification of data; WISP
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. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
Keywords
Computer aided instructionMachine learningRisk assessmentData disappearanceData protection impact assessmentsDPIAGDPRIllegitimate access to dataMachine learning applicationsMachine learning systemsPrivacyUnwanted modification of dataWISPData privacy
Links & artifacts
Suggested citation
Campanile, L., Forgione, F., Mastroianni, M., Palmiero, G., & Sanghez, C. (2022). Evaluating the Impact of Data Anonymization in a Machine Learning Application [Conference paper]. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 13380 LNCS, 389–400. https://doi.org/10.1007/978-3-031-10542-5_27