Visa's position as one of the world's largest payment network enables us to serve billions of consumers and tens of millions of merchants around the world on a daily basis. Our mission at the data analytics team is to unlock the hidden value in this massive amount of high-quality financial data, and develop a powerful commerce intelligence engine to drive innovations in the payment industry and beyond. We conduct applied research in machine learning and related fields to deliver unique insights and capabilities to our partners and clients and help them grow their business. Our research areas include deep learning applications in financial data, supervised and unsupervised representation learning, predictive modeling for fraud and risk management, consumer behavior modeling, recommendation systems, scalable machine learning systems, privacy preserving data analytics, and related fields.
- Deep learning for commerce – develop predictive models that are utilized by many parties in the payment and commerce ecosystem using advanced deep learning technologies.
- Scalable machine learning system – experiment with new generations of distributed systems for machine learning applications to meet the challenge of ever increasing scale and complexity in the data processing pipeline.
- Security analytics – enable proactive defense in our critical network and system infrastructure through continuous monitoring and real-time anomaly detection using machine learning technologies.
Cryptography forms the backbone of modern payment systems. At Visa Research, we are at the cutting edge of cryptographic research, moving the field forward by exploring its most exciting areas and research questions. These include provable security, zero-knowledge proofs, secure multiparty computation, post-quantum cryptography, and numerous other fundamental and applied aspects. Below are a few key areas of our focus:
- Secure Collaborative Computing and Privacy Enhancing Technologies – We design, implement, and evaluate protocols that enable parties who do not necessarily trust each other to participate in collaborative computations. Our protocols ensure that the parties' inputs remain private while accomplishing the shared objective, even when faced with malicious and corrupt participants. To achieve this, we utilize and propose advancements in various privacy-enhancing technologies, such as secure multiparty computation, zero-knowledge proofs, differential privacy, federated learning, and more.
- Post-Quantum Cryptography – We propose new cryptographic systems that can withstand attacks even from adversaries armed with fully operational quantum computers. Our research includes novel hardness assumptions, protocols, and deployment strategies, and we offer thought leadership to the payments industry in preparing for a world where quantum computers are a reality.
- Cryptography for Security Applications – We tailor and design cryptographic primitives necessary in security and privacy-allied areas such as computer networks, storage systems, identity management, blockchain, e-payment systems, and related technologies.
The Digital Currencies team is building the next generation of financial systems that rely on digital currencies, including decentralized cryptocurrencies like Bitcoin and Ethereum as well as semi-decentralized digital currencies like stablecoins and central bank digital currencies (CBDCs). Compared to traditional financial systems, these networks have significantly stronger resilience against cyberattacks thanks to blockchain protocols. This is primarily achieved by minimizing trust in various system components, i.e., even if part of these systems are taken over by an adversary and controlled maliciously, the system would still be able to function properly.
- On-chain scalability – sharding techniques allow consensus protocols to scale to a significantly larger number of nodes. The direct benefit of this is a considerably higher level of decentralization and thus a higher resiliency against cyberattacks.
- Off-chain scalability – cryptographic payment channels can break the scalability barrier of on-chain scalability and allow decentralized transaction processing to happen at massively high throughputs. Deploying them in a hub-and-spoke network topology with a decentralized deferred settlement model can potentially allow the payment system to match the level of what Visa can process today.
- Off-chain programmability – with the rise of decentralized finance (DeFi) applications that allow institutions to offer robust financial applications with significantly lower maintenance overheads but at a small scale, we believe off-chain programmability solutions would allow financial institutions to offer less expensive and significantly faster services by executing smart contracts off-chain, while still preserving the core benefits of decentralization through cryptographic off-chain protocols.
Identity and Authentication
Payment systems worldwide are gradually transforming to become ‘identity-first’. Consumers are increasingly using their mobile device, digital identity credentials, or biometrics to conduct payments. To enable Visa to be at the forefront of this transformation, the identity and authentication team invents, designs, and builds next-generation digital identity, authentication and biometric systems using cutting edge privacy, security, and machine learning techniques. The team’s vision is seamless, secure, privacy-preserving, transparent and intelligent authentication, and identity checks. The team’s expertise is highly interdisciplinary ranging from deep learning, federated learning, and adversarial learning to applied cryptography, biometrics, and security.
- Privacy-preserving Digital Identity – Identity credentials are being increasingly digitized worldwide, for example, mobile driver licenses in the US and electronic IDs in Europe. Furthermore, privacy regulations are being developed to regulate acceptance and use of identity credentials in each geography (for example, General Data Protection Regulation and California Consumer Privacy Act). The team is researching methods and techniques to enable digital identity credential ecosystem to be user-centric, decentralized and privacy-aware.
- Passwordless Authentication Security – User credentials such as PINs and passwords are the prevalent way of authentication on the web. However, use of such credentials is fraught with security, usability, and privacy issues such as phishing, and lost or stolen credentials. To overcome these limitations, an accelerating trend in industry is passwordless authentication. We are researching and developing methods and techniques for secure passwordless authentication.
- Interoperable and Secure Biometric Recognition – Biometric traits, both physiological (e.g., face, finger, iris, voice) and behavioral (e.g., gait, keystrokes) are being increasingly adopted for seamless, highly accurate and secure user recognition in payment applications. We are designing methods and techniques for large scale biometric systems to become interoperable, privacy-preserving, and end-to-end secure.
Quantum Information Sciences and Quantum Computation
As the size and reliability of quantum computers continue to increase, practical applications where they outperform classical computers are expected to emerge. Understanding their capabilities and impact in the payments industry is a central part of our mission. To that end, we study questions in quantum information processing and quantum computing, including:
- Exploring where quantum computers may have an edge over classical computers.
- Assessing which aspects of the payment ecosystem could be influenced by the introduction of quantum computers.
- Examining quantum cryptographic solutions to understand their advantages and benefits over classical alternatives.