The application provides an active
living body defense method and an electronic device, and belongs to the technical field of
living body detection. The method generates sparse
time domain keys by relying on transaction timestamps and a secure pseudo-random
number generator, randomly divides challenge probe frames and standard frames, and controls hardware registers in
layers to generate
time division multiplexing video streams. The
distortion frame timing cannot be predicted by attackers, and only a small number of frames are subjected to
optical distortion, and the collection process is non-invasive. Through
time domain physical response
verification and nonlinear timing overflow
verification double front-end screening, using the inherent defects of real hardware imaging rules and AI face changing model timing
smoothing, real-time deep fake videos can be efficiently identified. After
verification, the distorted probe frame is reversely optically reconstructed, and then the
double network is used to extract the micro identity features and
macro three-dimensional topological features of the face respectively, and the
living body is determined by comparing the
spatial mapping consistency of the two, forming a multi-layered and deep defense, and giving consideration to the security capabilities of non-invasive collection, high-precision identity verification and resistance to high-order AI face changing attacks.