Deep Learning/PUF-based Item Identification for Supply Chain Management in a Distributed Ledger Framework

Item identification is an important issue in modern supply chains where product items are traced by means of electronic tags that, despite the advantages they bring in terms of automation, are still subject to counterfeiting attacks such as tag modification, cloning, and re-application. To counter these threats (i) tags must be able to safely demonstrate their identity without giving attackers the ability to clone them or modify their content, and (ii) auxiliary identification techniques should be adopted in the case tags are removed from original items and reapplied to fake ones. Using Physically Unclonable Functions as tag 'digital fingerprint' has proven to be a viable solution for tag authentication. On the other hand, artificial intelligence has proven its effectiveness in the field of object identification. In this paper we describe a tag architecture immune to cloning and modification attacks, and illustrate how this can be coupled with a product item in a supply chain scenario to confer a stable and durable identity for authentication, in order to identify the items and track them in a distributed ledger framework. We also propose a deep learning approach to perform anti-counterfeiting controls when tag reapplication attacks are in place.