ID verification tags aren’t much use if someone can just peel them off and stick them to a fake product. MIT scientists have now designed ID tags that use the glue itself as a kind of fingerprint, and will scramble the barcode if someone peels it off.
The principle of the tag is the same as the radio frequency identification tags (RFIDs) commonly used to track stock and verify authenticity. Basically, each tag has a unique identifying code that can be read using a scanner to prove the item is the real deal. But in practice, all it’s technically doing is verifying that the tag is authentic, not whatever it’s stuck to. If anyone wanted to bypass the system, all they had to really do was peel off the tag and stick it on their fake item. So for the new study, the MIT team developed a new tag that would destroy its barcode if it was removed.
The trick is to embed the ID not in the tag itself, but in the glue sticking it to the item. Microscopic metal particles were mixed into the glue, then after it’s stuck on a surface, it’s scanned with high frequency terahertz waves. The metal particles bounce these waves back to the reader, taking a snapshot of how those metal particles have arranged themselves. That random pattern becomes like a fingerprint, which is used to identify the tag and is stored in the cloud. If someone tries to peel the tag off and restick it to something else, they’ll mess up that very specific arrangement of metal particles in the glue, returning the wrong ID when it’s later scanned.
“These metal particles are essentially like mirrors for terahertz waves,” said Ruonan Han, an author of the study. “If I spread a bunch of mirror pieces onto a surface and then shine light on that, depending on the orientation, size, and location of those mirrors, I would get a different reflected pattern. But if you peel the chip off and reattach it, you destroy that pattern.”
The team says the ID tag is small, measuring just 4 mm2 (0.006 in2), allowing it to be attached to a range of items, and is cheap enough to produce at scale. A machine learning model was trained to detect the patterns with more than 99% accuracy.
In its current form, the researchers say the system works with a sensor being up to 4 cm (1.6 in) away from the tag, and within a 10-degree angle. That would be fine for uses like scanning items in a warehouse, for example, but not so much for recognizing cars passing through a toll booth. The researchers plan to work towards addressing these shortcomings in future.
The research is due to be presented at the IEEE Solid States Circuits Conference in April.
Source: MIT