Privacy-aware BLE Proximity Detection through Optimizing Trade-offs

Abstract

In recent times, Bluetooth Low Energy (BLE) has emerged as a viable tool for proximity detection in diverse applications, including indoor positioning, social interaction monitoring, and asset management.

This thesis offers an multi-faceted study of BLE Received Signal Strength (RSS) proximity detection via measurement-, simulation-, and theoretical-based investigations. The thesis evaluates the efficacy and privacy trade-offs and provides valuable insights that may be used in the creation of systems that are both efficient and protect users’ privacy. To start, we provide an introduction of wireless positioning algorithms, emphasizing many benefits as well as the possible drawbacks, together with the opportunities the positioning technologies may have in the area of proximity detection. Thereafter, we take a look at the currently available BLE RSS proximitydetection approaches and the algorithms that lie behind them.

We introduce several performance metrics that may be used to compare different factors which may be odds with privacy preservation, e.g., accuracy and utility. To enhance the effectiveness of proximity detection, we offer an approach based on BLEthat overcomes possible privacy concerns related to social-interaction monitoring. We provide a detailed study of the privacy trade-offs associated with BLE systems since the privacy of the user is a significant concern in many of the applications that include proximity detection. While maintaining the accuracy of proximity detection, we identify potential risks and provide mitigation strategies to preserve the user privacy. In addition, we study the role that cryptographic methods and decentralized data storage play in preserving the privacy of user information while also preserving their privacy. However, the cryptographic-based methods are only briefly overviewed, as they do not form the core of this thesis. By utilizing privacy-preserving techniques, such as added noise, implementing obfuscation, and particularly a novel proposed argmax-based mechanisms, we achieve remarkably high levels of privacy in relation to building sizes, while maintaining detection probabilities around 90% and false-alarm probabilities below 15%.

Our study reveals that the opportunistic network for positioning can withstand a non-disclosure rate of up to 30% amongwearables without a substantial impact on positioning accuracy. However, as the non-disclosure rate exceeds 30%, the accuracy experiences a significant decline. Thus, it is important to maintain a disclosure rate below this threshold to ensure a satisfactory positioning performance.

Also, we conducted measurement campaigns at Tampere University (TAU) and University ’Politehnica’ of Bucharest (UPB) to investigate various sources of BLE errors. First,we showthat the impact of hardware includes variations in the median RSS up to 5 dB among devices of identical models due to antenna characteristics, calibration difficulties, and differences in receiver sensitivities. Second,we showthat the BLE advertising channel index affects the positioning accuracy, as deploying all channels led to an increased standard deviation, especially in scenarios involving the setup at TAU. Third, we also show that environmental characteristics play a role, as evident by the difference in standard deviation between datasets (at both univeristies’ premises) even under Line-of-Sight conditions. Fourth, the hardware orientation contributes to RSS variability, with notable differences observed across different angles. Finally, the co-existence of 2.4 GHz band devices, such as WiFi, affects BLE systems, as deactivatingWiFi resulted in a mean RSS increase of 2.5 dB, suggesting potential interactions in the chipset.

To sum up, this dissertation contributes with an outlook on BLE RSS-based proximity-detection methods by designing a privacy-aware algorithm showing good performance in digital contact-tracing scenarios and by conducting empirical and theoretical studies to evaluate privacy–utility/accuracy trade-offs. We believe that our findings provide valuable insights that could be used in the design, implementation, and the vast adoption of the upcoming proximity-detection methods that could achieve a reasonable trade-off between performance and privacy preservation.