Silicon Twins: MIT’s New Chip Fabrication Method Enables Unforgeable Hardware Authentication Without External Servers

MIT engineers develop a method for two chips to share a unique fingerprint. Learn how "twin PUFs" eliminate the need for external server authentication.

By: AXL Media

Published: Feb 26, 2026, 8:43 AM EST

Source: The information in this article was sourced from MIT News

The Vulnerability of Digital Fingerprints

Every CMOS (Complementary Metal-Oxide-Semiconductor) chip produced today possesses a unique identifier known as a Physical Unclonable Function (PUF). These are the result of unavoidable, microscopic manufacturing variations—effectively, a digital fingerprint for hardware. While these IDs are nearly impossible to replicate, traditional cryptographic schemes require this "fingerprint" data to be stored on an external server for later verification. This storage creates a central point of failure: if the server is breached, the hardware security is compromised. On February 20, 2026, MIT researchers announced a way to bypass this risk by creating "twin" chips that share the same secret from birth.

[Image showing a large, detailed view of a tiny silicon chip with grids and holes]

The 'Torn Paper' Analogy

To understand the novelty of the MIT method, lead author Eunseok Lee suggests a simple analogy: imagine tearing a sheet of paper in half. The jagged edges of the tear are random and impossible to replicate, but the two halves fit together perfectly. While silicon wafers cannot be "torn" in this way, the MIT team developed a process to share randomness across the boundary of two adjacent chips before they are diced. This ensures that the two resulting chips have an identical, shared PUF that is unique only to that specific pair. This allows one chip to directly "unlock" the other without any third-party mediator.

Harnessing Transistor Breakdown