SecureFind: Securing the Crowdsourced Lost-and-Found via Stealth Polling

TL;DR

SecureFind is a novel framework that turns the ubiquity of smartphones into a massive, private search party for lost items. By combining BLE tags with a "dummy traffic" mechanism, it ensures that only the object owner discovers the item's location, preventing curious "finders" from stealing the prize and keeping participant locations anonymous.

The "Curious Finder" Paradox

Crowdsourcing the recovery of lost items (like children, pets, or high-value electronics) is efficient but inherently risky. If a service provider or a crowdsourced "detector" knows exactly where a high-value item is, the incentive to report it is often outweighed by the incentive to keep it. This is the Object Security problem. Simultaneously, participants don't want a central server tracking their every move to see if they are near a lost item—the Location Privacy problem.

Existing solutions like Tile or Apple’s Find My network have made strides, but often rely on centralized trust. SecureFind targets a zero-trust environment where the service provider is "honest-but-curious."

Methodology: The Art of Hiding in Plain Sight

SecureFind utilizes a multi-round polling process based on the Framed Slotted ALOHA protocol.

1. The Basic Scheme: Maximum Security

In the basic version, every participant in a target area acts as a "potential detector." When a search is initiated, participants who don't see the lost tag generate dummy responses. To the server and other detectors, these dummies look exactly like a "hit" on the lost object. Only the owner, who knows the unique ID of the tag and the random seeds used for polling, can filter out the noise to find the real signal.

2. The Advanced Scheme: Selected Polling

While the basic scheme is secure, it is communication-heavy. The authors introduce Selected Polling, where only a subset of time slots ($\omega$) are reported.

• The Statistical Shield: The system uses Pearson's chi-squared test to ensure that the distribution of "1s" (hits) reported by the real detector is statistically indistinguishable from the "1s" reported by fake detectors.
• Dynamic Adjustment: The owner adjusts the number of real positions queried per round based on previous results, maximizing efficiency without crossing the threshold of "detectability."

Note: The workflow involves the Owner initiating a request to the Service Provider, who then tasks Mobile Detectors to scan for the BLE tag using randomized bit vectors.

Experimental Validation

The authors conducted extensive simulations with 10,000 detectors over a 16 $k{m}^2$ area.

• Recovery Probability: With modern urban densities, the probability of an object being within range of at least one detector is over 90%.
• Security vs. Efficiency: The Advanced Scheme (AS) reduced detector-to-server communication by nearly 80% compared to the Basic Scheme (BS).
• Energy Impact: A single search operation consumes less than $1.5imes10^{-3}$ J—totally negligible compared to a smartphone's 20,000 J battery capacity.

Note: The results demonstrate that as the p-value threshold ($p_{thre}$) increases, the "Real Detector's Rank" moves toward 0.5, indicating perfect anonymity.

Critical Insight: Why This Matters

The genius of SecureFind lies in its Inductive Bias toward privacy. Instead of trying to encrypt the location (which is difficult during a live search), it obfuscates the existence of a match.

Limitations & Future Work

• Collusion: The current model assumes the Service Provider does not collude with detectors. If they did, the dummy response mechanism might be vulnerable to sophisticated traffic analysis.
• Incentives: While the paper assumes an incentive structure exists, the "selfishness" of users in a real-world deployment (e.g., turning off Bluetooth to save energy) remains a hurdle for coverage.

Conclusion

SecureFind bridges the gap between massive-scale crowdsourcing and individual privacy. By turning every participant into a potential "dummy," it creates a shield of plausible deniability that protects the item, the detector, and the owner.