High-precision continuous identity authentication is crucial for preventing privacy breaches and safeguarding sensitive data in the digital intelligence era, garnering substantial interest from both academia and industry. However, the complexity of identity verification and the need for unperceived sensing limit the availability of suitable methods. This study proposes a novel approach using tri-axial plantar load distributions, visualized through heatmaps to highlight inter-subject feature variations and the complementary effects of multi-axial components. Identity authentication experiments, conducted on a 20-subject dataset across straight walking and turning motions using a one-dimensional convolutional neural network (1D-CNN) classifier, achieved the error rate of only 1.03%, significantly outperforming single-axis pressure sensing (3.63%). To our knowledge, this is the first study to employ tri-axial plantar load as a mechanical signature for identity authentication, which is expected to advance technologies in fields such as wearable devices and biometric security systems.