Two Smartphones Steal 3D Printer IP via Side Channels at 98.89% Accuracy

98.89% command-level G-code reconstruction accuracy from two smartphones 60 cm away, non-line-of-sight, using only acoustic and magnetic emissions. That’s the number researchers just published on arXiv, and it kills the assumption that 3D printer IP theft requires expensive equipment or a direct line of sight.

Dual Sensors Beat Single-Mode Attack Limits

Previous side-channel attacks on 3D printers had three problems: low reconstruction accuracy, short effective range, and reliance on obvious lab gear like laser microphones or high-end magnetometers. This work uses the microphones and magnetometers already inside two off-the-shelf smartphones. The team placed them 60 cm from the printer with no direct line of sight — think behind a desk or around a corner. One phone records the acoustic signature of stepper motors and extruder movements; the other captures the magnetic field fluctuations from the same motors. Fusing both modalities pushes command-level reconstruction to that 98.89% figure.

Transferability Confirms the Threat Isn't a Fluke

The researchers didn’t stop at one printer in one room. They repeated the attack on a different 3D printer in a different environment — different brand, different table, different background noise. The attack still worked, which means the side-channel signatures are generic enough to make G-code reconstruction practical across machines. Reconstructed G-code gives you the full toolpath, layer heights, extrusion rates — the entire IP baked into the print.

Commodity Hardware Exposes a Structural Weakness

Two smartphones you already own can now walk into a makerspace or factory floor and walk out with a complete copy of any printed object’s production instructions. No physical tampering, no network intrusion, no line-of-sight required. The paper is a clear signal that additive manufacturing security needs physical-layer countermeasures, not just network firewalls. Acoustic dampening, magnetic shielding, or intentional signal masking on the printer side will become table stakes for any facility printing proprietary parts.

Source: A Non-Line-of-Sight, Multi-Modality-based Side-Channel IP Theft Attack on Additive Manufacturing Using Dual Smartphones
Domain: arxiv.org