Vladimir Balygin, Svyatoslav Kushnarev
In the current era of the internet’s fast development, businesses and governments are collecting a broad range of data about people from browsing history and shopping preferences to personal information. This spike in data collection has led to individuals being flooded with targeted advertising and, in some cases, risking their privacy, security and personal lives.
Recognising the critical need for data protection, many countries have adopted privacy laws inspired by the European Union General Data Protection Regulation (EUGDPR) – one of the toughest privacy laws at the present time. Thus, the issue of data safekeeping has become increasingly pressing in places where privacy laws mandate businesses to demonstrate their customers’ data is secure.
Modern privacy-enhancing data protection mechanisms implement secure defences to protecusers from data breaches, theft, and misuse. Users benefit from these technologies through effective data sharing capabilities and strong defenses against potential abuse. A significant challenge for businesses is to demonstrate that their privacy enhancing technology complies with any prevailing privacy laws, in particular by demonstrating that their implemented protections provide authentic defenses.
In this context, among various privacy defenses that have been proposed, differential privacy and its extensions, notably “d-privacy,” stand out the most. This approach uses randomisation and probabilistic methods to protect users’ data and is employed by technological giants such as Apple, Google, agencies of the U.S. government, etc.
More recently d-privacy has been enhanced by secure shuffling mechanisms, which have been claimed to provide more protection for users in various applications, including electoral voting, online auctions, secure communication, and privacy-preserving data analysis. This poster explores the mathematical concepts of secure shuffling and its role in reinforcing the fundamental attributes of d-privacy: utility, privacy, and verifiability.