『聯合KYT』商業計畫書隱私強化技術FHE賦能跨金融業交易監控

Abstract

隨著詐騙與洗錢手法日益科技化與跨境化，單一金融機構在交易監控上常面臨「單兵作戰、見樹不見林」的困境，難以有效攔阻跨媒介金流跳轉與人頭帳戶的濫用行為。本研究提出一項創新解決方案：「以隱私強化技術FHE（Fully Homomorphic Encryption，全同態加密）賦能的聯合KYT（Know Your Transaction）資訊平台」，使金融機構能在不揭露明文個資的前提下，於密文中共享風險模型，聯合進行異常交易監控。 研究首先系統梳理國內外反洗錢與個資保護法規，確認隱私運算於現行法規下具備可合規的發展空間；進一步規劃資訊平台技術架構、模型部署方式與 API 訂閱式商業模式，並與玉山金控「金融無塵室」洽談策略合作，進行技術互補之驗證及中長期合作藍圖規劃。技術面上，本團隊採用通用GPU加速TFHE運算，開發混合密文資料格式與風險特徵對齊模組，顯著提升AI推論效率與系統的可商轉性。 「聯合KYT」資訊平台預期將為金融業提供一套兼具隱私保護與跨機構聯防能力的交易監控解方，透過技術落地、商業模式設計與政策連結，實現「防詐攔阻」、「民眾財損降低」與「隱私安全」的三贏目標。中期願景為協力推動「金融無塵室2.0」的商轉應用，進而促成主管機關、金融公會與金控銀行共組「聯偵中心JKIC（Joint KYT FHE&AI Consortium）」，建構具隱私強化功能的交易監控聯合中心。 最後，亦建議將FHE賦能金融業的成功經驗，進一步擴展至醫療產業，推動跨院共享AI模型之隱私運算平台。

As fraud and money laundering techniques become increasingly sophisticated and transnational, individual financial institutions face significant limitations in transaction monitoring—akin to “fighting alone and missing the forest for the trees.” It is therefore challenging to effectively detect cross-channel fund transfers and prevent the exploitation of mule accounts. In response, this study proposes an innovative solution: a Joint KYT (Know Your Transaction) platform powered by Fully Homomorphic Encryption (FHE), enabling financial institutions to collaboratively monitor suspicious transactions and share risk models within encrypted data, without disclosing any personally identifiable information (PII). The study begins by reviewing relevant anti-money laundering (AML) and data protection regulations, both domestic and international, to confirm the regulatory feasibility of privacy-preserving computation. It then designs the platform architecture, model deployment strategy, and API-based subscription business model. A strategic collaboration with E.SUN Financial Holding’s “Financial Clean Room” is proposed to verify technical compatibility and co-develop a partnership roadmap. On the technical front, the team utilizes general-purpose GPUs to accelerate TFHE computation and develops hybrid ciphertext data formats and feature alignment modules to significantly enhance AI inference efficiency and system deployability. The Joint KYT platform is expected to provide the financial sector with a privacy-preserving, cross-institutional solution for transaction monitoring. Through practical implementation, business model innovation, and regulatory alignment, the platform aims to achieve a mutually beneficial outcome across fraud prevention, financial protection, and privacy preservation. In the medium term, the vision is to support the commercialization of “Financial Clean Room 2.0,” and to initiate the formation of the Joint KYT FHE&AI Consortium (JKIC)—a collaborative detection center uniting regulators, financial industry associations, and financial holding companies. Furthermore, the experience of empowering the financial industry through FHE is suggested as a blueprint for developing privacy-preserving AI model sharing platforms across hospitals in the healthcare sector.