Our Online World Relies on Encryption. What Happens If It Fails?
Quantum computing taps into unusual properties of the very small—where particles can exist in multiple states simultaneously (quantum superposition) and stay connected over distance (entanglement), allowing a quantum computer to explore many possibilities at once, significantly speeding up certain computations.
“Our approach is expected to be inherently resistant to both classical and quantum attacks,” says Ruckenstein. “It would not only strengthen public confidence in AI systems, but also unlock new opportunities for data-powered, socially responsible innovation.”
Protecting Data During Use
Modern encryption methods, developed roughly 50 years ago, could not envision the computational demands of today—let alone those of the quantum era. Relying on hard-to-solve mathematical problems, these systems mostly only protect data in transit or at rest—leaving it exposed during use. That poses a problem for data-intensive applications like AI training models, which process vast amounts of data that is often private or confidential. Current approaches typically require models to decrypt data during training, leaving it exposed, or employ privacy-preserving techniques that slow processing speeds, making them difficult to apply at scale.
The BU-led NSF project offers a new path forward. The proposed scheme, called Encrypted Operator Computing (EOC), merges physics, computer science, and mathematics to develop scalable methods for computing directly on encrypted data—long considered the “holy grail” of cryptography.
“The approach is an alternative to Fully Homomorphic Encryption (FHE), an elegant, state-of-the-art cryptographic tool, which has so far proven difficult to apply to large-scale practical problems,” says Ruckenstein.
The EOC allows users to manipulate and gain insights from confidential data without ever exposing the raw information to third parties. This level of security and privacy is essential for applications such as blockchain transactions, medical AI models, cloud services, and more.
“While our EOC method is designed to work on classical computers doing classical computations, the conceptual breakthrough behind it is quantum computation–inspired,” says Claudio Chamon, a CAS professor of physics. “In addressing the real-world challenge of computation on encrypted data, we also encounter fundamental questions, such as how many distinct ways a given computation can be expressed for a fixed-length circuit. We relate these questions to thermodynamic concepts like ‘entropy,’ which describes how unpredictable or random a system is based on how many ways it can be arranged.”
How entropy applies to computation is the subject of the team’s PNAS paper.
“In our framework, computation is represented as a circuit of logical elements, or gates, encoding elementary operations and which, when applied sequentially to the input data, implement the desired computation,” says Ruckenstein, who coauthored the paper with Chamon, Ran Canetti, a CAS professor of computer science, and Eduardo R. Mucciolo, a professor of physics at the University of Central Florida. “In the paper, we considered both the functionality and complexity of computational circuits—what the circuit is computing and how large a circuit is needed to implement that computation.”
The team’s physics-inspired approach treats complexity in computing as a thermodynamic quantity; thermodynamics relates to how things like heat and energy spread. The rules of thermodynamics govern how, for example, the heat diffuses in your morning coffee: as the heat spreads, the molecules become more distributed and disordered, their patterns more complex. None of which stops you from enjoying your coffee—but good luck recovering the history of all those erratic molecules. The researchers suggest that, in a computer circuit, its gates can similarly be disordered to hide information.
The PNAS paper proposes a dynamic process to obfuscate, or “hide,” any circuit by rearranging gates, randomizing its structure without altering its function. The aim is to not only scramble information fast, but also do so thoroughly—essentially destroying all patterns, so that a program is impossible to reverse engineer. The team’s vision is to create a trustworthy environment where both the data and programs that use the data stay hidden.
______________________________________________________
Program obfuscation is an extremely powerful and versatile concept for protecting data, its processing, and its various uses in multiple scenarios and over time. However, it is notoriously hard to construct: to date, we have no general-purpose program obfuscation scheme that is even close to being practical. This exciting project has the potential to make program obfuscation a reality.
— Ran Canetti
______________________________________________________
“Program obfuscation is an extremely powerful and versatile concept for protecting data, its processing, and its various uses in multiple scenarios and over time,” says Canetti. “However, it is notoriously hard to construct: to date, we have no general-purpose program obfuscation scheme that is even close to being practical. This exciting project has the potential to make program obfuscation a reality.”
Breaking Boundaries Through Convergent Research
The NSF-funded project aims to turn these cryptography concepts into practical tools. Together, the research team will develop the EOC framework into scalable, special-purpose hardware, merging physics-inspired insights about information with advanced cryptography and pure mathematics. The goal is to accelerate performance and make secure, privacy-preserving computing widely accessible for real-world use.
“By combining expertise from diverse areas, we can tackle problems from multiple angles at once—whether it’s understanding quantum behavior, designing new algorithms, or building better hardware,” says Mucciolo. “This synergy not only speeds things up, but also allows us to dive much deeper than any one discipline could alone. We’re uncovering connections that wouldn’t be visible without this kind of cross-disciplinary perspective.”
One of the team’s other contributors, Timothy Riley, a professor of mathematics at Cornell University, says the collaboration across disciplines is a “rare and precious opportunity” that is allowing the researchers “to understand each other’s languages, to learn from each other’s perspectives, and to share the models, problems, and abstractions that drive our work.”
Canetti, Chamon, and Ruckenstein were able to advance the work with the support of the Hariri Institute’s Quantum Convergence Focused Research Program, which facilitates convergent thinking and multidisciplinary collaborations across BU on crosscutting themes around quantum science and engineering. All three BU researchers are affiliated with the institute.
“The rise of digital infrastructure demands stronger security to protect our economy, privacy, and national interests,” says Yannis Paschalidis, a BU College of Engineering Distinguished Professor of Engineering, director of the Hariri Institute, and a member of the University’s Task Force on Convergent Research and Education. “Solving these complex challenges requires breaking down silos. This work shows how convergent research can drive real-world impact and unlock entirely new technological frontiers.”
Maureen Stnton is Assistant Director of Marketing & Communications, Hariri Institute for Computing, Boston University. The article was originally posted to the website of Boston University.
