July 2022 Summaries

The financial industry is increasingly dependent on information sharing among institutions to effectively combat financial crimes such as fraud and anti-money laundering (AML), a challenge highlighted by the Royal United Services Institute's Future of Financial Intelligence Sharing Research (FFIS) program. The study led by Nick Maxwell, "Lessons in private-private financial information sharing to detect and disrupt crime," explores how financial institutions can collaborate to share data while preserving privacy, which is crucial given the prevalence of customers engaging with multiple financial providers. With the average American and millennials having numerous financial relationships, individual institutions often lack a comprehensive view of a customer's activities, complicating efforts to identify illicit transactions. Innovations in technology, such as platforms like Duality, which combine cryptography and data science, are emerging to address these challenges, although they must navigate the constraints of regulations like the GDPR, which emphasizes data minimalism over the data-maximalist approach favored in AML efforts. For information sharing to become a standard practice, strategic vision and regulatory support are needed to balance effective crime detection with privacy considerations, as demonstrated by technology solutions that enhance data security while addressing non-technical issues.

Duality Technologies, recognized for its leadership in privacy-preserving data collaboration, has been included in the AIFinTech100 list for 2022, which highlights the most innovative AI and machine learning solution providers in financial services. The rapid adoption of AI and data analytics in the financial sector, driven by technological advancements and evolving regulations, offers significant cost-saving potential estimated at $447 billion over the next year. Duality's CEO, Dr. Alon Kaufman, emphasizes the importance of secure data collaboration in unlocking the value of sensitive data while maintaining privacy and confidentiality, a capability that is increasingly crucial for financial institutions. FinTech Global director Richard Sachar stresses the need for established institutions to stay informed about leading AI technologies to remain competitive. Duality, founded by renowned cryptographers and data scientists, specializes in privacy-enhancing technologies that enable AI on encrypted data while ensuring compliance and IP protection, earning recognition from organizations such as Gartner and the World Economic Forum.

Synthetic data refers to artificially generated information that mimics the structure and patterns of real-world data but does not link to actual individuals, offering privacy benefits over traditional anonymization techniques. It is increasingly used to train AI systems, correct biases in datasets, and expedite the development of proofs of concept, with Gartner predicting that by 2024, a significant portion of AI training data will be synthetic. Although synthetic data can closely replicate the accuracy of real data without posing re-identification risks, it also presents challenges, such as the inability to trace back to real-life individuals for practical problem-solving or exposing competitive insights. Despite these challenges, synthetic data is seen as a promising tool for enhancing data privacy and utility, particularly when combined with advanced techniques like homomorphic encryption, offering new opportunities for collaboration across industries while maintaining data security.

Duality Technologies has launched OpenFHE, a groundbreaking open-source fully homomorphic encryption (FHE) library developed in collaboration with major entities such as Intel, Samsung, the University of California – San Diego, and MIT. This advanced library aims to revolutionize data privacy by allowing computations on encrypted data without the need to decrypt it or share secret keys, thus accelerating the adoption of Privacy Enhanced Technologies (PETs). OpenFHE, a successor to PALISADE and other projects like HElib and HEAAN, offers post-quantum security and supports features like bootstrapping and scheme switching, as well as hardware acceleration through a standard Hardware Abstraction Layer (HAL). It addresses the usability issues of existing FHE libraries by providing a unified API and a user-friendly interface, enabling organizations to integrate FHE into their applications seamlessly. The project, supported by DARPA's financial backing, promises to enhance encrypted data computations with increased usability and security, leveraging Duality's expertise in secure data collaboration.

Duplicate invoice financing, a form of trade finance fraud, poses significant challenges in the banking industry, where businesses might take the same purchase order to multiple banks for financing due to the paper-based nature of the process and limited technological solutions. Current systems allow banks to make rudimentary checks, but regulations often prevent deeper inquiries, especially across borders, complicating efforts to identify and mitigate these fraudulent practices. Financial institutions face the dilemma of protecting relationships with longstanding customers while needing to verify invoice authenticity with other banks. Duality offers a solution by utilizing privacy-enhancing technologies that allow banks to collaborate without exposing sensitive information or violating regulations like GDPR, as data remains encrypted and does not leave the data center. This approach not only meets regulatory compliance requirements but also provides a secure way for banks to share insights and reduce the risks associated with duplicate trade finance. While initial concerns about the effectiveness and regulatory compliance of such technology existed, it is now recognized by technology and security leaders in large financial institutions as a valid method for protecting data while facilitating necessary collaboration.

Ian Quah shares his journey and experiences in exploring homomorphic encryption libraries, particularly highlighting his transition from a machine learning background to privacy-preserving machine learning (PPML) and eventually to homomorphic encryption. Initially finding it challenging to grasp the intricacies of Fully Homomorphic Encryption (FHE) due to a lack of cryptographic knowledge, Quah sought a library that was accessible and Python-compatible, leading him to OpenFHE, formerly known as PALISADE. He emphasizes the importance of several criteria when choosing a homomorphic encryption library, including the availability of information, community size and engagement, ease of use, open-source nature, and compliance with security standards. Quah eventually became part of the OpenFHE team, contributing to documentation improvements and library restructuring to enhance user learning experiences. He briefly discusses other libraries like Concrete, SEAL, and Lattigo, noting their strengths and drawbacks, and encourages potential users to consider joining the OpenFHE community while also recommending resources for further understanding of homomorphic encryption and its standards.

The blog post discusses the importance of standardization and peer review in establishing trust in open-source libraries, particularly in the context of Fully Homomorphic Encryption (FHE), a cutting-edge technology allowing computations on encrypted data without decryption. Since its proposal in 2009, FHE has undergone significant development and commercialization, necessitating public trust for widespread acceptance. Standardization, led by industry experts and organizations like homomorphicencryption.org, has been crucial, with recent contributions from the International Association for Standardization (ISO). Peer review and open-source publication are equally vital, ensuring security and transparency, exemplified by Duality's role in developing the PALISADE and OpenFHE libraries with collaboration from MIT, Intel, and Samsung. These efforts enable users to verify the implementation's correctness, fostering confidence in FHE's reliability, as demonstrated by its adoption by select Duality customers.