Abstract
We present a concept of physically unclonable functions utilizing the photoacoustic effect to generate structurally random, inference-resistant cryptographic keys. The system consists of a CuO/SnO₂ nanoparticle composite, where CuO acts as a visible-range absorber and SnO₂ serves as a non-absorbing dispersive matrix. Nanosecond laser pulses induce localized heating and acoustic wave emission, providing spatially heterogeneous photoacoustic signals that are digitized into binary matrices. Evaluations across ten devices yielded a bit uniformity of 49.54%, inter-device Hamming distance of 49.69%, entropy of 0.983, and bit aliasing of 49.38%-all approaching ideal values for secure key generation. Machine learning attacks using logistic regression and support vector machines failed to infer underlying patterns, with prediction accuracies of 53.53% and 52.54%. The device maintains cryptographic performance after transfer to diverse substrates, including human skin, highlighting its mechanical adaptability. This subsurface, light-to-sound-based approach offers a scalable platform for secure authentication on flexible or opaque surfaces.