Analyzing the potential of Differential Privacy in Data Sharing among Small and Medium-sized Enterprises
Authors
Knudsen, Jens William ; Koniaris, Iasonas
Term
4. term
Publication year
2022
Submitted on
2022-09-28
Abstract
Denne rapport undersøger, om differential privacy kan gøre datadeling mellem små og mellemstore virksomheder (SMV’er) mere sikker og tillidsfuld. Udgangspunktet er, at SMV’er sjældent deler data på grund af begrænset kapacitet, regulering som GDPR, og bekymringer om privatliv og fortrolighed. Projektet følger en analytisk tilgang med en state-of-the-art-gennemgang af privatlivsteknikker (anonymisering, k-anonymitet og differential privacy), kvalitative semistrukturerede interviews for at forstå praksis og bekymringer, samt brug af Diffusion of Innovations og netværkseffekter til at belyse mulig udbredelse og værdiskabelse. Den tekniske analyse adresserer centrale aspekter af differential privacy, herunder balancen mellem privatliv og nytte, injektion af støj, håndtering af gentagne forespørgsler, følsomhedsbinding, valg af privatlivsparameteren ε, modstand mod koblingsangreb, tillid, tilgængelige biblioteker og teknologisk modenhed. Rapporten sammenligner k-anonymitet med differential privacy, diskuterer mulige systemarkitekturer og giver anbefalinger og forslag til fremtidigt arbejde. Arbejdet er teoretisk og inkluderer ingen kodeimplementering. Uddraget indeholder ikke de endelige konklusioner, men indikerer et fokus på både fordele og udfordringer ved differentielt privat datadeling for SMV’er.
This report examines whether differential privacy can enable safer, more trustworthy data sharing among small and medium-sized enterprises (SMEs). It starts from the observation that SMEs share data infrequently due to limited resources, regulations such as the GDPR, and concerns around privacy and confidentiality. The project adopts an analytical approach that combines a state-of-the-art review of privacy techniques (anonymization, k-anonymity, and differential privacy), qualitative semi-structured interviews to capture practice and concerns, and the use of Diffusion of Innovations and network effects to consider adoption dynamics and potential value. The technical analysis covers key elements of differential privacy, including the privacy–utility trade-off, noise injection, repeat-query risks, bounding sensitivity, choosing the privacy parameter ε, resistance to linkage attacks, trust, available libraries, and technology readiness. The report compares k-anonymity with differential privacy, discusses possible architectures, and presents recommendations and directions for future work. The study is theoretical and does not include code implementations. The excerpt does not provide final findings, but indicates a focus on both benefits and challenges of differentially private data sharing for SMEs.
[This summary has been generated with the help of AI directly from the project (PDF)]
Documents