Reliable anti-counterfeiting checks in extreme conditions

Researchers at the National University of Singapore (UXO) have invented a new anti-counterfeiting method called DeepKey. Developed in just eight months, this security innovation uses two-dimensional (2D) material labels and artificial intelligence-enabled (AI) authentication software.

Compared to conventional anti-counterfeiting technologies, DeepKey works faster, achieves very accurate results and uses permanent identification tags that are not easily damaged by the environment such as extreme temperatures, chemical spills, exposure to UV radiation and moisture. This new authentication technology can be applied to a variety of high-value products, from medicines, jewelry and electronics. For example, DeepKey is suitable for labeling COVID-19 vaccines to allow rapid and reliable authentication, as some of these vaccines need to be stored at an ultra-cold temperature of -70 ° C.

Under the leadership of doc. Prof. Chen Po-Yena and doc. Prof. Wang Xiaonan of the Department of Chemical and Biomolecular Engineering at UXO Technical School, the team’s 2D material safety labels show physically inflexible function patterns (PUF patterns), which are randomly generated by creasing thin films of 2D material. Complex patterns of 2D multi-scale material can be classified and validated by a well-trained deep learning model, enabling reliable (100 percent accuracy) authentication in less than 3.5 minutes.

Existing anti-counterfeiting technologies using PUF samples typically face several bottlenecks, including complicated manufacturing, a specialized and laborious reading process, long authentication times, insufficient environmental stability, and costly fabrication.

“With this research, we have addressed several bottlenecks encountered by other techniques,” said doc. Prof. Wang. “Our 2D PUF tags are environmentally stable, easy to read, simple and inexpensive to make. In particular, the adoption of deep learning has significantly accelerated overall authentication, pushing our invention a step further into practical application.”

The researchers published their results in a scientific journal The thing On December 2, 2020, this research was conducted in collaboration with researchers from Anhui University of Technology and Nanyang University of Technology.

Stable, simple and scalable procedure for creating PUF tags

Exceptionally, researchers do not need special equipment to create secure labels. They can be easily made with a balloon, a bottle with a 2D dispersion of material and a brush.

“First we inflate the balloon and brush on its surface with viscous ink for 2D material. After air drying overnight, we inflate the balloon. Due to the mutual mechanical mismatch of 2D material and latex substrate, large surface, crumpled PUF Samples are generated during contraction. These samples PUFs can be subsequently cut to the required size, and usually hundreds can be made at once, ”said Dr. Jing Lin, a member of the research team.

Further, the researchers quickly record the PUF tag with an electron microscope, which is then synchronized with their innovative software to go through a process of deep learning of “classification and validation”. “The whole procedure takes less than 3.5 minutes, most of which are waiting to be read from an electron microscope. The authentication itself is very fast, in less than 20 seconds,” Dr. Jing explained.

Fast authentication using AI algorithms for deep learning

All PUF key-based technologies have extremely high coding capacities due to the huge number of different patterns that can theoretically be produced. However, the large coding capacity also leads to a long time of authentication, as the “search and compare” sample check has to be performed in a huge database. This trade-off between high encryption capacity and long authentication time often limits such anti-counterfeiting PUF tags from practical applications.

“With our new technology, we are breaking this long-standing trade-off between high encryption capacity and long authentication time using classified 2D-material PUF tags and deep learning algorithms,” said doc. Prof. Wang.

First, the researchers used various 2D materials to produce PUF tags with recognizable AI characteristics. Second, they trained a deep learning model to implement a two-step authentication mechanism. “We used the deep learning model to pre-categorize PUF patterns into subgroups, so the search and comparison algorithm is implemented in a much smaller database, which shortens the total authentication time,” explained doc. Prof. Wang.

Currently, the only technologies available are similar to this UXO innovation polymer-based labels. Wrinkled polymer markings are confirmed based on surface samples, just like the new 2D material markings. However, their authentication currently requires individual extraction and matching of features, which is slow and shows a reliability of only 80 percent. The authentication of the UXO team is enhanced by deep learning, and thus it is much faster and reaches almost 100 percent validation accuracy.

In addition, in relation to the wet chemical preparation of boron-based polymer labels, which includes the use of harmful organic chemicals and UV light, the technique of making UXO researchers is much faster and safer.

Next steps

The UXO team has applied for a patent for their invention and now plans to take this technology a step further. “We are looking for better, faster and more robust read and authentication approaches for our PUF tags,” said doc. Prof. Wang.

The team has already begun conducting research on other reading techniques to further shorten processing time. “In addition, such naturally encoded information with PUF tags can be further ensured by staying on the blockchain, so that the entire supply chain and quality control can be transparently monitored,” added prof. Wang.


Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of the news published on EurekAlert! contribution to institutions or for the use of any information through the EurekAlert system.