Collaborative mining across multiple data owners is valuable, but centralizing transactional data is often infeasible due to privacy and policy constraints. This paper presents FedFHUIM, a federated workflow for frequent high-utility itemset mining in a client-server setting where data remain at clients and only pattern-level summaries are exchanged. Each client performs local mining using thresholds derived from the global support and utility requirements and shares candidate pattern summaries with the server. The server merges the received candidates and performs an exact verification step by collecting local support and utility contributions for the merged candidates, thereby recovering globally valid frequent high-utility itemsets without transferring raw transactions. We implement the workflow using a lightweight socket-based architecture and evaluate it on benchmark utility datasets under multiple client counts. The results show an exact match with centralized mining in the tested settings and demonstrate practical runtime and memory behavior for moderate client counts. At higher client counts, relaxed local thresholds can increase candidate volume and verification cost, indicating a scalability trade-off. These findings support the feasibility of exact federated high-utility mining and motivate future work on tighter candidate filtering and evaluation under heterogeneous client distributions.