DPNetTraffic - Compartilhamento de dados de tráfego de rede utilizando privacidade diferencial

• Ano de defesa: 2023
• Autor(a) principal: Monteiro, Felipe Cavalcante
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Privacidade de dados; Privacidade diferencial; Tráfego de redes; Pós-processamento; Data privacy; Differential privacy; Network praffic; Post-processing; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://repositorio.ufc.br/handle/riufc/76093

Network traffic data is extremely valuable and plays a fundamental role in a variety of applications. This information is collected by different entities, such as Internet service providers (ISPs), who often share or commercialize this data with external entities. However, sharing this data can potentially compromise the privacy of individuals whose information is contained within it. To address this privacy concern, we propose a new approach called DPNetTraffic. This approach utilizes the concept of differential privacy, which involves adding a controlled level of noise to the original data. This preserves the privacy of useful information in the data, as the shared data does not reveal specific and identifiable information about them. The DPNetTraffic approach stands out for its efficiency in preserving the privacy of useful information of network traffic data. Compared to other techniques that also adopt differential privacy, experimental results have shown that it introduces less noise to the data. This means that valuable insights and useful information can be obtained from the shared data while protecting the privacy of the individuals involved. The use of DPNetTraffic in sharing network traffic data represents a significant advancement in balancing data utility and user privacy protection. This approach has the potential to be adopted by entities that collect and share network traffic data, offering a reliable and effective solution to mitigate privacy breaches. In comparison to the specific baseline results, the proposed approach, named DPNetTraffic, demonstrated significantly superior performance. An average reduction of 35% in the introduced noise was observed in the data.