The current wave of Industrial Internet of Things (IIoT) is reshaping the manufacturing sector with system interoperability, remote real-time process monitoring and advanced analytics. As digitally enabled manufacturing machines continue to grow exponentially, it becomes imperative to uniquely and securely identify them in the cyber-physical world, particularly in defense, biomedical, energy and aerospace manufacturing. Research about threats originating from internal adversaries’ i.e the machine/organization owner within a tiered digitally connected supply chain is scarce. This paper introduces a machine fingerprinting scheme named as the ‘Witness Box Protocol’ (WBP) that exploits the physical properties of manufacturing machines (legacy or smart) and their surroundings to create a unique biometric like fingerprint. WBP provides both machine registration and authentication on a digital network through a low cost, non-invasive approach. The fingerprint is generated by a Locality Sensitive Hashing (LSH) technique that accommodates small variations in physical signature data and can corroborate data provenance from machines by verifying machine identity through authentication. Additionally, this fingerprint hash simplifies asset management within a large enterprise or distributed network comprising of thousands of machines in a Manufacturing-as-a-Service (MaaS) paradigm. In this research, fingerprints were randomly generated from the statistical features of signals from 3D printers and CNC machine in a production-like lab environment. Using k-means clustering and Jaccard similarity index, these fingerprints are shown to identify the source equipment with 95 % accuracy.

当前的工业物联网（IIoT）浪潮正在通过系统互操作性，远程实时过程监控和高级分析重塑制造业。随着数字化制造机器的继续呈指数级增长，在网络物理世界中，特别是在国防，生物医学，能源和航空航天制造领域，必须唯一，安全地标识它们。对于来自内部对手（即分层数字连接的供应链中的机器/组织所有者）的威胁的研究很少。本文介绍了一种称为“见证箱协议”（WBP）的机器指纹识别方案，该方案利用制造机器（旧式或智能型）及其周围环境的物理特性来创建独特的生物识别特征，例如指纹。WBP通过低成本，非侵入性的方法在数字网络上提供了机器注册和身份验证。指纹是通过本地敏感哈希（LSH）技术生成的，该技术可适应物理签名数据中的细微变化，并可以通过通过身份验证来验证机器身份，从而确认来自机器的数据出处。此外，此指纹哈希简化了大型企业或分布式网络中的资产管理，该大型企业或分布式网络包含“服务即制造（MaaS）”范例中的数千台机器。在这项研究中，在类似生产环境的实验室环境中，根据3D打印机和CNC机器的信号统计特征随机生成了指纹。使用k均值聚类和Jaccard相似度指数，