A few years ago, MIT researchers invented a cryptographic ID tag that is several times smaller and significantly cheaper than traditional radio frequency tags (RFIDs) that are often affixed to products to verify their authenticity.

This smaller tag, which offers improved security over RFIDs, uses terahertz waves, which are smaller and travel faster than radio waves. But this terahertz tag shared a major security weakness with traditional RFIDs: A counterfeiter could peel off the original item’s tag and reattach it to a fake one, and the authentication system would be none the wiser.

Researchers have now overcome this security risk by exploiting terahertz waves to develop an anti-tampering ID tag that still offers the advantages of being small, cheap and secure.

They mix microscopic metal particles into the glue that sticks the tag to an object, and then use terahertz waves to detect the unique patterns that form on the particles’ surface. Similar to a fingerprint, this random glow pattern is used to authenticate an item, explains Eunseok Lee, an Electrical Engineering and Computer Science (EECS) graduate student and lead author of a paper on anti-tampering tags.

“These metal particles are basically like mirrors for terahertz waves. If I spread a bunch of mirror pieces on a surface and then shine light on it, depending on the orientation, size and location of those mirrors, I get A different reflection pattern will be obtained. But if you peel off the chip and reattach it, you destroy the pattern,” says Rowan Hahn, an associate professor at EECS who researches electronics. He leads the Terahertz Integrated Electronics Group at the laboratory.

The researchers developed a light-activated anti-tampering tag with a size of about 4 square millimeters. They also demonstrated a machine learning model that helps detect tampering by identifying similar glow pattern fingerprints with over 99 percent accuracy.

Because the terahertz tag is so cheap to produce, it can be implemented in a large supply chain. And its small size enables the tag to be attached to objects too small for traditional RFIDs, such as certain medical devices.

The paper, to be presented at the IEEE Solid State Circuits Conference, is a collaboration between Hahn’s group and the Energy Efficient Circuits and Systems Group of Anantha P. Chandrakasan, MIT’s Chief Innovation and Strategy Officer and Dean of the MIT School of Engineering. , and Vaniver Bush Professor of EECS. Co-authors include EECS graduate students Zebi Chen, Maitri Ashok, and Jaewon Won.

Prevention of tampering

This research project was partly inspired by Han’s favorite car wash. The business stuck an RFID tag on his windshield to validate his car wash membership. For added security, the tag was made of fragile paper so it would be destroyed if a less-than-scrupulous customer tried to peel it off and stick it on a different windshield.

But this is not a very reliable way to prevent tampering. For example, one can use a solution to dissolve the glue and safely remove the fragile tag.

Instead of validating a tag, a better security solution is to authenticate the item itself, Hahn says. To achieve this, the researchers targeted glue at the interface between the tag and the surface of the object.

Their anti-tampering tag has a series of tiny slots that enable terahertz waves to pass through the tag and strike microscopic metal particles that are embedded in the glue.

Terahertz waves are small enough to detect particles, while larger radio waves are not sensitive enough to see them. In addition, using terahertz waves with a 1-mm wavelength allowed the researchers to create a chip that did not require large, off-chip antennas.

After passing through the tag and striking the surface of the object, the terahertz waves are reflected, or backscattered, to a receiver for verification. How those waves are backscattered depends on the distribution of the metal particles that reflect them.

The researchers added multiple slots to the chip so that the waves could strike different points on the surface of the object, gaining more information about the random distribution of particles.

“These responses are impossible to replicate, unless the Glow interface is destroyed by the counterfeiter,” Hahn says.

A vendor will take initial readings of an anti-tampering tag when it’s stuck on an item, and then store that data in the cloud, using it later for authentication.

AI for verification

But when it came time to test the anti-tampering tag, Lee ran into a problem: It was too difficult and time-consuming to measure accurately enough to determine if two glue patterns matched.

He reached out to a friend at the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) and together they tackled the problem using AI. They trained a machine learning model that could compare glow patterns and calculate their similarity with over 99 percent accuracy.

“One drawback is that we had a limited sample of data for this demonstration, but we can improve the neural network in the future if a larger number of these tags are deployed in the supply chain,” says Lee. , which gave us a lot of data samples,” says Lee. .

The authentication system is also limited by the fact that terahertz waves are highly lossy during transmission, so the sensor can only be 4 cm from the tag to get an accurate reading. This distance would not be a problem for an application such as barcode scanning, but would be too short for some potential uses, such as in an automated highway toll booth. Also, the angle between the sensor and the tag must be less than 10 degrees or the terahertz signal will drop too much.

Hahn says he plans to address these limitations in future work, and hopes to make other researchers more optimistic about what can be achieved with terahertz waves, despite the many technical challenges. Will be encouraged to be.

“One thing we really want to show here is that the application of terahertz spectrum can go beyond broadband wireless. In this case, you can use terahertz for ID, security and authentication. There are so many possibilities,” he added.

This work was supported in part by the US National Science Foundation and the Korea Foundation for Advanced Studies.