[Paper Review] Short Paper: Privacy Comparison of Contact Tracing Mobile Applications for COVID-19

Mohamed‐Lamine Messai, Hamida Seba|arXiv (Cornell University)|Oct 7, 2020

COVID-19 Digital Contact Tracing7 references4 citations

TL;DR

This paper evaluates and compares the privacy properties of major contact tracing mobile applications for COVID-19, focusing on Bluetooth and GPS-based systems. It analyzes how design choices—particularly centralized vs. decentralized architectures—affect user privacy, concluding that decentralized, open-source models with strong anonymization and data deletion offer superior privacy protection compared to centralized or GPS-dependent systems with weak or no privacy safeguards.

ABSTRACT

With the COVID-19 pandemic, quarantines took place across the globe. In the aim of stopping or slowing the progression of the COVID-19 contamination, many countries have deployed a contact tracing system to notify persons that be in contact with a COVID-positive person. The contact tracing system is implemented in a mobile application and leverages technologies such as Bluetooth to trace interactions between persons. This paper discusses different smart-phone applications based on contact tracing system from privacy point of view.

Motivation & Objective

• To analyze and compare the privacy risks inherent in various contact tracing mobile applications deployed during the COVID-19 pandemic.
• To identify the primary sources of privacy threats in these applications, including health authorities, malicious developers, infected users, and hackers.
• To evaluate how different technical designs—especially centralized vs. decentralized architectures—affect user privacy.
• To assess the role of anonymization, source code transparency, and data retention policies in enhancing or undermining privacy.

Proposed method

• The study conducts a comparative analysis of 8 major contact tracing applications, including TraceTogether, Aarogya Setu, AC19, StopCovid, Stopp Corona, Covid-Watch, PACT, and DP-3T.
• Applications are evaluated based on predefined privacy criteria: use of Bluetooth or GPS, need for a trusted server, support for false claims, anonymization techniques, openness to source code, and data destruction policies.
• The analysis focuses on threat vectors such as deanonymization by health authorities, surveillance by malicious developers, and data breaches by hackers.
• The paper categorizes systems into centralized (server-dependent) and decentralized (peer-to-peer) models, assessing their respective privacy trade-offs.
• Each application is scored on key privacy dimensions, including whether it uses random identifiers, stores data on servers, or deletes data after a set period.
• The comparison is structured using a table that maps each app to its privacy-relevant properties, enabling direct contrast across design choices.

Experimental results

Research questions

• RQ1 How do centralized versus decentralized contact tracing architectures differ in their privacy implications for users?
• RQ2 To what extent do anonymization techniques and open-source status reduce privacy risks in contact tracing applications?
• RQ3 What are the main privacy threats posed by health authorities, malicious developers, infected users, and hackers in these systems?
• RQ4 How do GPS-based and Bluetooth-based systems compare in terms of privacy leakage and user tracking potential?
• RQ5 What role does data retention and deletion play in minimizing long-term privacy exposure?

Key findings

• Centralized applications like Aarogya Setu and AC19, which rely on a trusted server and use GPS or unencrypted data, pose higher privacy risks due to centralized data storage and lack of anonymization.
• Decentralized systems such as PACT and Covid-Watch, which do not require a central server and use anonymous identifiers, significantly reduce the risk of mass surveillance and data breaches.
• Open-source applications like TraceTogether and Stopp Corona demonstrate better trustworthiness and transparency, reducing the risk of hidden tracking mechanisms by developers.
• Applications that implement strong anonymization and delete data after a defined period—such as StopCovid and Stopp Corona—offer stronger privacy guarantees than those without such policies.
• The use of GPS in applications like AC19 and Aarogya Setu increases privacy leakage and battery consumption, making them less suitable for long-term, large-scale deployment.
• Even in systems with anonymization, users may still infer infection status through notification receipt, indicating an unavoidable privacy loss in the contact tracing model itself.