Jess, SimonSimonJessIhrig, LukasLukasIhrigKrause, ArturArturKrauseMesserschmidt, ClaudiusClaudiusMesserschmidtFörster, FelixFelixFörsterBursac, NikolaNikolaBursac2026-07-032026-07-032026-0336th CIRP Design Conference, CIRP Design 2026https://hdl.handle.net/11420/63742Modern product development increasingly relies on data-driven approaches to make informed decisions. Especially within the framework of System Generation Engineering (SGE), where new products are based on reference systems, and thereby a valuable knowledge base is generated. Preserving the sovereignty of this data, particularly in the context of product designs, is crucial. Simultaneously, there is a strong desire to leverage this data for data-driven services that provide valuable predictions, such as production costs, processing times, production risk, or the Product Carbon Footprint (PCF). To resolve this conflict of interest, encryption techniques offer a solution. This study investigates the practical applicability of homomorphic encryption (HE): Using PCF prediction as a case study, it demonstrates that machine learning models can forecast such values. Subsequently, it examines whether this is also feasible during inference with homomorphically encrypted data. To this end, a neural network trained on real-world industrial data is compared with a model that performs computations on encrypted data. The results demonstrate that prediction quality decreases only slightly when using HE, confirming its suitability for preserving information quality. However, this advantage comes with a significant performance overhead: computation times for the encrypted calculations increase substantially, posing a challenge for practical application, especially in time-critical scenarios. The paper concludes that HE is a promising approach to reconcile data sovereignty with data-driven innovation. However, its broad industrial application crucially depends on future advancements in technology’s computational efficiency and scalability.en2212-827120268590Elsevierhttps://creativecommons.org/licenses/by-nc-nd/4.0/Technology::620: EngineeringJournal Articlehttps://doi.org/10.15480/882.1740910.1016/j.procir.2026.05.22810.15480/882.17409