[Paper Review] Mobile Privacy-Preserving Crowdsourced Data Collection in the Smart City

Joshua Joy, Ciarán McGoldrick+1 more|arXiv (Cornell University)|Jul 11, 2016

Privacy-Preserving Technologies in Data19 references3 citations

TL;DR

This paper proposes Authorized Analytics, a scalable, privacy-preserving architecture for crowdsourced IoT data collection in smart cities, where mobile devices (e.g., autonomous vehicles) locally perturb their data using a two-coin randomized response mechanism to ensure strong local differential privacy. The system achieves high utility with relative error below 0.5% in large-scale deployments (1M devices) and enables secure, decentralized stream analytics without synchronization or centralized trust.

ABSTRACT

Smart cities rely on dynamic and real-time data to enable smart urban applications such as intelligent transport and epidemics detection. However, the streaming of big data from IoT devices, especially from mobile platforms like pedestrians and cars, raises significant privacy concerns. Future autonomous vehicles will generate, collect and consume significant volumes of data to be utilized in delivering safe and efficient transportation solutions. The sensed data will, inherently, contain personally identifiable and attributable information - both external (other vehicles, environmental) and internal (driver, passengers, devices). The autonomous vehicles are connected to the infrastructure cloud (e.g., Amazon), the edge cloud, and also the mobile cloud (vehicle to vehicle). Clearly these different entities must co-operate and interoperate in a timely fashion when routing and transferring the highly dynamic data. In order to maximise the availability and utility of the sensed data, stakeholders must have confidence that the data they transmit, receive, aggregate and reason on is appropriately secured and protected throughout. There are many different metaphors for providing end-to-end security for data exchanges, but they commonly require a management and control sidechannel. This work proposes a scalable smart city privacy-preserving architecture named Authorized Analytics that enables each node (e.g. vehicle) to divulge (contextually) local privatised data. Authorized Analytics is shown to scale gracefully to IoT scope deployments.

Motivation & Objective

• To address privacy risks in large-scale, real-time crowdsourced data collection from mobile IoT devices in smart cities.
• To enable secure, privacy-preserving data sharing without relying on centralized trust or synchronization.
• To balance utility and privacy in dynamic, distributed IoT environments such as autonomous vehicle networks.
• To provide a shared, extensible privacy infrastructure for researchers deploying real-world smart city experiments.

Proposed method

• Each mobile device independently applies a two-coin randomized response mechanism to its sensitive data before transmission.
• The first coin determines whether to report the true value (with probability p) or a random value (with probability 1−p).
• The second coin, flipped only when the first is tails, produces a randomized response with probability q.
• The aggregator estimates the true count of 'yes' responses using the formula: $ Y_E = \frac{Y_R - (1-p) \cdot q \cdot N}{p} $.
• The system operates without synchronization, enabling scalable, real-time stream analytics across millions of devices.
• Local data perturbation ensures strong plausible deniability and prevents re-identification of individual contributors.

Experimental results

Research questions

• RQ1 How can mobile crowdsourced data be collected in smart cities while preserving strong local differential privacy?
• RQ2 What is the optimal balance between privacy and utility in decentralized, large-scale IoT data streams?
• RQ3 Can a decentralized, synchronization-free system achieve high utility and strong privacy guarantees simultaneously?
• RQ4 How does system performance scale with increasing numbers of devices and varying query frequencies?

Key findings

• The system achieves a relative error of less than 0.5% in query results when 1 million devices answer every 10 seconds, with results computed every second.
• Utility improves significantly with increasing numbers of devices due to the law of large numbers, reducing relative error.
• Reducing the query execution interval improves utility by increasing batch size, which enhances estimation accuracy.
• The system maintains strong local privacy guarantees without requiring centralized trust or synchronization among devices.
• The two-coin randomized response mechanism provides a superior balance between utility and privacy compared to prior approaches.
• The architecture scales gracefully to IoT-scale deployments and supports real-time stream analytics in untrusted, distributed environments.