Data science is a key field for revealing "what's next," and data scientists have a lot of powerful, sophisticated tools for turning data into critical insights. Often what they don't have is access to high-value data stores they can use to refine their models and make their breakthroughs. In machine learning, better tools and better data produce better outcomes; but data privacy regulations can keep data scientists from rich, fresh sources of insight and optimal results.
This is especially true in industries like financial services, health and life sciences, and government and public service, where data sensitivity is high and the data is protected by law. To benefit from the value of that private data without violating privacy or confidentiality, many organizations are adopting a new approach called encrypted learning.
Encrypted learning is a machine learning technique that trains an algorithm on encrypted data without decrypting it. Data privacy and security are protected by default, and so data scientists can now use previously unavailable private data to improve the accuracy of their models. Here are some ways different organizations can use encrypted learning to improve their data models and, in turn, deliver greater value to their organizations and customers.
Financial transactions are a potentially rich source of information. Every credit card purchase, payroll deposit, and bank transfer is associated with detailed personal information that is protected by regulation. If that information were available, it could provide tremendous value by revealing trends and habits.
By gaining access to this new source of data with encrypted learning, financial services firms can obtain a more precise understanding of those transactions at a level of detail that was previously not possible. Richer, more precise data can be used to debias old financial models, better identify trends, and unlock millions of dollars in previously unrealized business value through a more granular understanding of consumer spending habits. It could also be used to identify and prevent losses associated with financial fraud, or better manage investment portfolios.
Data modelling in healthcare is tricky. Researchers know the potential for improved patient care and better outcomes exists, but medical records are extremely sensitive, and regulations like the Health Insurance Portability and Accountability Act (HIPAA) make using that data difficult. And because patient safety is a top concern, training algorithmic models on incomplete datasets is not ideal.
Encrypted learning can help medical researchers solve that problem. As an example, an international life sciences firm can access large sets of specific diagnostic data, including data from jurisdictions around the world, to develop a cancer screening system that promises faster, more precise diagnoses to support pathologists with automated test analysis. Encrypted learning can also allow pharmaceutical researchers to share data without compromising privacy or intellectual property in order to accelerate drug discovery in response to health crises.
Government agencies collect a lot of private data from the citizens they serve. Nearly every aspect of life in modern society involves sharing personal, health, financial, and other data with a public agency at the local, state, or federal level. Most times, government agencies are bound by strict regulation from sharing that data with their colleagues in other areas of government—even though that data could be used to improve services in vital areas such as health and safety.
Today, government organizations could use encrypted learning to share high-value, protected data to facilitate interagency cooperation, and develop and refine risk models and improve services. The insights resulting from these cooperative efforts are being used to enable a variety of use cases. Some examples include a better understanding of trends affecting emerging public health crises, better protection for children in the care of child welfare services, and more efficient allocation of public resources in areas affected by crime, poverty, and climate change.
Researchers and data scientists recognize the potential that private, protected data has to train machine learning algorithms to support better decision-making. Encrypted learning holds the key to unlock that value.
To learn more about encrypted learning and how protected, high-value data can be used safely to improve business processes, register for the Cape Privacy webinar. Hear Priceline CTO, Martin Brodbeck and Cape Privacy CEO, Che Wijesinghe discuss important topics in data science, data privacy, and machine learning. They will share key insights into encrypted learning and modern privacy-enhancing technologies, as they relate to digital commerce, consumer goods, financial services, healthcare, and media. Get the latest from the front lines of data science.