Brief

Researchers have demonstrated that Active Motor Noise Cancellation (AMNC), a hardware defense against acoustic side-channel attacks in 3D printers, is ineffective against vibration-based attacks. While AMNC successfully neutralizes the acoustic channel, the vibration channel still leaks information about the printed object. A temporal model analyzing the vibration waveform over the course of printing achieved approximately 61% accuracy in identifying printed objects, indicating a significant vulnerability that AMNC does not address. The study utilized data from Bambu Lab printers and found the vibration leak to be device-specific, suggesting that while AMNC protects the acoustic channel, other channels like vibration, magnetic, or power remain susceptible to attacks. AI

Researchers have developed LLM-ADAM, a novel framework utilizing Large Language Models for anomaly detection in additive manufacturing G-code files. This system decomposes the task into distinct roles: an Extractor-LLM to structure process parameters, a Reference-LLM to interpret documentation, and a Judge-LLM to identify deviations. Evaluated on a corpus of 200 FFF G-code files, the best configuration achieved 87.5% accuracy in detecting defects like under-extrusion and warping, significantly outperforming a baseline model. AI