MIT engineers embed unclonable 'fingerprints' directly into secure chips
Engineers at MIT have created a new way to securely authenticate chips using tiny physical differences in their structure. The method turns these variations into unique 'fingerprints,' known as physical unclonable functions (PUFs). Unlike traditional security measures, this approach keeps secret keys locked inside the chip itself, removing the need for external storage.
The process starts during chip production, where engineers carefully adjust transistor gate oxides along the edges of wafers. These adjustments create pairs of adjoining transistors on two nearby chips with matching but distinct breakdown states. Using low-cost LEDs, the team triggers these states, generating shared secret keys that never leave the chip.
Because the keys are built into the hardware, no external memory or additional encryption layers are required. This reduces both the risk of attacks and the power needed for secure communication. The method also avoids expensive equipment, relying instead on standard LEDs and existing CMOS manufacturing processes.
The technology, called Twin-PUF, allows secure data exchange between devices without extra controllers. While still in early development, it shows promise for simplifying security in Internet of Things (IoT) systems. So far, testing has focused on medical sensors and implantable devices, but broader applications remain unexplored.
This advancement could cut down on the complexity and vulnerabilities in IoT security. By embedding unique signatures directly into chips, the method removes the need for external key storage. The approach is also scalable, as it fits into current semiconductor production lines without major changes.
