Tag-based tracking ecosystems help users find lost keys and luggage, but the same infrastructure can be turned against people for stalking. Every deployed defense — Apple’s unwanted tracking alerts, the Google/Apple DULT specification, and prior academic detectors — rests on the same assumption: that a rogue tag exposes a stable identifier, or at least beacons in a predictable way. That assumption does not survive contact with an adversary who simply rotates identifiers aggressively. We show that identifier-based defenses break down entirely against such trackers, and that a physical-layer property closes the gap.

Analog imprints don’t rotate

Logical identifiers can change arbitrarily fast; the transmitter’s analog imprint cannot. AirCatch is a passive detector built on this asymmetry. A rogue tag that hides behind thousands of fresh identifiers still lands in the same compact region of Carrier Frequency Offset (CFO) feature space, and keeps occupying it for as long as it follows the victim.

AirCatch contributes along three axes:

• a modulation-aware CFO fingerprint that augments packet-level CFO with content-independent CFO components, amplifying device distinctiveness
• a detection algorithm based on core density and persistence, robust to contamination and to evasion, using per-identifier segmentation
• BlePhasyr, a ≈$10 BLE SDR built from commodity components, making RF-fingerprinting-based detection practical in resource-constrained deployments

Evaluation

We evaluate across Apple, Google, Tile, and Samsung tag families in multi-hour captures. Evasion is stress-tested systematically with a scenario generator sweeping a grid of transmission and rotation periods, and validated in real-world mobility traces — home and office commutes, public transport, car travel, and airport transit — while sweeping background tag density.

Key results

• No false positives across the full range of stress tests
• Early detection over a wide span of adversarial configurations and environments
• Graceful degradation confined to extreme low-rate regimes, which also reduce the attacker’s own tracking utility

Why it matters

Identifier rotation is cheap for an attacker and, until now, decisive. AirCatch moves detection to a layer the attacker cannot rewrite in firmware, and does so on hardware costing about the price of lunch. Defenses against unwanted tracking should stop assuming the tracker will identify itself.

Paper

📄 Paper appearing in PETS’27.