Idemia granted patents for document fraud detection and image enhancement
Two U.S. patents with potential identity and biometrics implications, one issued April 21 for a “Method For Determining A Point Spread Function of An Imaging System,” and one published earlier in the month for a “Method For Detecting Document Fraud,” have been awarded to Idemia.
The latest patent for a method to detect document fraud involves using two images; a “first image of a first document and a second image of a second document” in question. According to the patent, “A procedure of detection of zones sensitive to document frauds are applied in the regions of the first image and of the second image registered on the first image. Each sensitive zone detected is then divided into a plurality of subparts. A measurement of dissimilarity is calculated between corresponding subparts from the first image and the registered second image. It is then determined whether the first document is identical to the second document from measurements of dissimilarity.”
So, if the first document is different from the second document, “a level of difference is determined between the first and second documents according to a value representing a proportion of different subparts; and fraud is detected when the level of difference is below a third predetermined threshold,” says the patent description.
Many fraud attacks are based on falsified documents, the patent states, noting for example that this is the case “with intrusion methods based on falsified identity documents” and “with frauds consisting of opening multiple bank accounts in online banks offering bonuses when accounts are opened” explaining that “the detection of document frauds is, therefore, an important challenge in terms of security, but also in economic terms.”
“Some document fraud techniques consist of locally modifying an original document to produce a new document. For example, these techniques may consist of replacing a photograph in a document or modifying at least one letter or figure in a text area of said document. The document obtained is very similar to the original document but contains sufficient differences locally to defraud,” the inventors, Thibault Berger, Laurent Rostaing and Alain Rouh stated in the patent description
They said while there are techniques that exist “for preventing such frauds, some are based on insertion in the documents of security patterns or unfalsifiable information such as biometric information in passports,” which is “encoded in an electronic chip of some passports. However, these techniques require either having physical access to the document or using a specific imaging system making it possible typically to produce images in a near-infrared spectrum or in ultraviolet fluorescence, which is not possible for online procedures where a user uses, for example, a smartphone to transmit only one or more color images of his documents.”
One identified solution, they said, “consists of carrying out a classification of a digitized version of a document to know the type thereof (identity card, car registration document, driving license, passport),” an that “once the document is classified, high-level information is extracted, such as text fields … such as optical character recognition fields … or information representing a face contained in the document. Prior classification facilitates the extraction of relevant information for fraud detection. It is known where to seek this information in the digitized document. In general, these solutions have the drawback of being specific to the document being studied and are sensitive to algorithm errors in the extraction of high-level information.”
The paper, Document Fraud Detection at The Border: Preliminary Observations on Human and Machine Performance, presents a reasonably “broad outline of approaches for detecting document fraud in the case of border control,” the inventors stated, but added that “the automatic methods [they’ve] proposed… are based mainly on an analysis of specific security patterns as well as on the formation of a reference document base, which is often a weak point,” they asserted.
But “another approach,” the inventors posit, “consists of calculating a digital signature of an image of the document and saving it. This saved signature can then be used subsequently to authenticate another image of this same document. This approach, which is found in the patent document FR 3047688, requires an enrolment of each document in order to calculate its reference signature. This approach cannot, therefore, be used if this enrolment and a suitable infrastructure for managing the signatures are not available. The use of signatures makes it possible to determine whether two documents are identical or different but does not make it possible to determine whether the differences are due to a fraud.”
They explained that this approach “is desirable to overcome these drawbacks of the prior art.”
They further state that it is particularly “desirable to propose a method making it possible, without prior knowledge of characteristics of the documents and without any prerequisite with regard to the digitisation of said documents, to determine whether a first document is the result of one or more local modifications of a second document, which would represent an attempt at document fraud.”
The inventors explained in their approved patent that “The method of the invention is completely generic since it requires no a priori knowledge about the documents analysed, about the devices that acquired the images of the document, about the angle of photographing of the documents and about the illumination of the documents at the moment of photographing.”
According to “one embodiment,” they pointed out, “the registration procedure comprises determining that no fraud has been detected when a value representing an efficacy of matching of points of interest between the first and second images is below a predetermined threshold … thus the registration method makes it possible to reject documents of different types.”
The patent approved on April 21, Method For Determining A Point Spread Function Of An Imaging System, is described as “a method for determining a point spread function of an imaging system, a method for improving an image acquired by an imaging system and a device implementing at least one of said methods.”
“For each position in a plurality of positions of a target in an optical field of the imaging system,” the process involves “acquiring an image of the target, referred to as the real image; obtaining a synthetic image of the target representing a digital model of the target adjusted to a zone of the real image corresponding to the target so that the model coincides with the zone; estimating the point spread function using the real image and the synthetic image; and calculating an optical transfer function by applying a Fourier transform to the point spread function,” then “calculating an average optical transfer function from the optical transfer functions calculated for each position in the plurality of positions; and obtaining an average point spread function by applying an inverse Fourier transform to the average optical transfer function.”
The context of the invention involves “a point spread function (PSF), also referred to as an optical pulse response (OPR) … a mathematical function describing a response of an imaging system to a point source. The PSF models how a point is imaged by an imaging system and reflects optical deformations of the imaging system. It is considered that an image of an object issuing from an optical system is the result of a convolution between a signal issuing from the object (i.e. an image issuing from a perfect imaging system without optical deformation) and the PSF modelling the imaging system. When this function is known, it is possible to deconvolution (i.e. to apply an inverse convolution) the image issuing from the imaging system by the PSF in order to obtain an improved image of the object.”
The method “uses a target, in the form of a test grid, to determine the PSF. This test grid is placed facing the lens system, at a position corresponding to a focal distance of said system, this focal distance being dependent on the size of the test grid. This method therefore allows only to determine the PSF corresponding to the position of the test grid. The PSF obtained is therefore not valid for any other position of the test grid. An image of an object situated at a position other than the position for which the PSF was calculated could not be correctly corrected by the PSF. This method involves calculating another PSF for this other position and having means for determining the position of the object with respect to the imaging system.”
In other words, objects are identified and analyzed for enhancement, such as through generating a synthetic image of the object by adjusting it, or geometrically transforming it.