Physical Unclonable Functions (PUFs) are expected to overcome a number of constraints mainly encountered in the security of the Internet of Things (IoT) domain. They are supposed to be an effective approach to protect objects against counterfeiting by providing unique identities to them through applications such as authentication and key generation in digital environments. To evaluate their security potential, it is essential to quantify their randomness that is intimately related to the fabrication process. It is the reason why previous works were mainly motivated by developing solutions that improved the randomness without inducing overhead on other concerns such as cost, power consumption and integrability. The present work focuses on the diversification capability offered by complex patterned oxide thin films obtained through Chemical Beam Vapour Deposition (CBVD) to be used as PUF and discusses the relevance of this solution from a conceptual point of view. A major advantage consists in the possibility to benefit from a large range of different physico-chemical properties that can additionally be evaluated by a large panel of different characterization techniques, that provide huge encoding capacity as they evaluate to different randomly distributed values at each point. As with all constructions of PUF, the challenge is then to define appropriate additional blocks to optimize the uniformity, reproducibility, and uniqueness of the PUF responses. The concept of a fully miniaturized and integrated optical PUF based on the thin film fingerprints is described, as very promising solution.