Defects are regarded as entities that have to be removed for better performance in most of materials. However, in recent years, interests in controlling the defect structures itself of liquid crystals are increasing due to their unique properties. Defect engineering with liquid crystals have been applied to various scaffold for lithography, soft actuators, controlling living matter, and scaffold for self-assembly. The defects of the liquid crystals have various structures such as point, ring, and line according to the molecular arrangement around the defect cores. Here, we fabricated the complex defects arrays by using the appropriate confinement effect and the intrinsic elasticity of the liquid crystal. These defects act as seeds for the assembly of chiral liquid crystals and assemble them in a random pattern with a lattice. These assembled patterns memorize the randomness or natures because their formation is governed by thermal fluctuations. We showed that these arbitrary patterns can be applied on a physical unclonable function (PUF), similar to artificial fingerprints, by using object detector based on deep learning. In addition, they have reconfigurability that is not possible with silicon-based PUF because this self-assembly is fully. This new authentication tag based on self-assembly is expected to be able to prevent attacks on personal information effectively in the coming 4th industrial revolution era.