Thwarting forgery with paper fingerprints

While on the surface a sheet of paper may seem like any other sheet, manufacturing paper is a complex process and each sheet is unique.

Verifying official documents
“Typically, wooden particles are used as the base and multiple substances are used to stick these particles together,” explains lead author Ehsan Toreini, a Ph.D. student in Cyber Security in Newcastle University’s School of Computing Science.

“Our idea was that the majority of paper used for official and legal documents, certificates, invoices and so on is not completely opaque. Different types let through different levels of light and reflect it in different ways and as a result, each one reveals a unique fingerprint.

“We proposed an algorithm that generates a compact and unique identifier for each sheet of paper. This identifier is then converted into a QR code which can be verified efficiently off-line by anyone.

“Since this identifier is basically representative of that paper texture, any illegal modifications - including copying the contents of the document to another paper sheet - could be identified.”

In order to make each sample – or fingerprint - from the paper sheet consistent, the team used a rectangular box (dimensions 37 mm x 57 mm) and asked users to only capture what was inside this box with the camera.

“Our algorithm recognizes this box automatically and the texture patterns inside this area are analyzed for the fingerprint. We only need one sample from this area to construct its fingerprint,” says Toreini.

How the system might be used
This technique can be applied to prevent counterfeiting of any physical document that is made of paper material, for example, e-passports and bank notes. One potential application is to verify the authenticity of a university degree certificate. Before the certificate is issued, the paper texture is extracted from the certificate, digitally signed by the institution and printed on the certificate as a barcode.

Anyone wanting to verify the authenticity of the certificate would just need to take a picture of the texture fingerprint and the system will compare the new image with the digitally signed copy stored in the barcode.

The algorithm can also correctly identify the box and automatically justify it to the desired position and condition by rotating and masking the undesired parts mostly caused by human errors. This makes it robust against human mistakes while taking the photo.

“A potential employer could easily authenticate the degree certificate by themselves without having to contact the university.” says Dr Hao.

— Read more in Ehsan Toreini et al., “Texture to the Rescue: Practical Paper Fingerprinting based on

Texture Patterns,” ACM Transactions on Information and System Security (25 May 2017)