Reliable identity(ID) is the cornerstone of product verification for supply chain trust and security. Traditional popular ID techniques, such as serials numbers and bar codes, can be either easily cloned. While novel ID techniques, such as the ones based on physical unclonable functions (PUFs) or nano-chemicals, exploit the randomness in micro/nano-scale to make the cloning difficult. However, the PUF-based ID is applicable only on electronics and the nano-chemical ID usually requires either specific fabrications or inconvenient verification processes. To address these shortcomings, we propose a novel 3D ID tag with random micro- structure features, preventing the cloning through the technical difficulties of the structure reproducing. The proposed ID can be produced in a cost efficient way, applied on most products with a solid surface, and verified conveniently by cell-phone level equipment. In this paper, we introduce the design and the fabrication of our tag. Then we take the tag's optical images and develop the verification algorithm enhanced by a machine-learning-based object detection. Experimental results from our 3D tag prototypes demonstrate the reliability of the verification.