Biometric-Like Approach for Verifying Artworks Authenticity

The artwork market is plenty of unauthorized reproduction of original products. One of the most varies filed is the counterfeiting of Authenticity Certificate related to paints, lithography, sculptures, etc., with the aim to create an illegal market of reproduced copies. To resolve this problematic, it is possible change the current paper certificate, related to a single artwork, with a digital version, which will contain some specific information, related to the artwork itself. In this paper, starting with the well-known advantages given by the biometry paradigm in human authentication, we propose a method able to distinguish the single “non-living” objects. In other words, we propose an approach that, by using the random inimitably characteristics, is able to uniquely identify artworks such as painting, lithographs, sculptures, etc. In this way it could be possible creating a secure digital certificate of authenticity (digital COA). Due to the high density information available in modern acquisition media, it is possible using a Speckle Metrology approach. During verification phase, the same area has to be acquired, to extract embedded verification data. It is possible to secure this data using a private key, necessary for accepting the digital signature. The presence of possible geometrical distortions between image present in the certificate and acquired during the verification phase, it is necessary applying geometrical corrections based on affine transformation, before executing the correlation methodologies, used in speckle metrology.


Introduction
When we purchase a work of art, the main problem is to obtain a genuine certificate of authenticity (COA).
There is a great misuse of "certificate of authenticity"; if the COA is not directly originated by the artist, it is pretty much meaningless.In any case, even if the certificate of authenticity shall be made by the artist, in general, is relatively simple to clone it.Moreover, the certificate of authenticity is very often directly made by the seller.
In a simple, a dishonest seller can duplicate both the certificate of authenticity that the work of art.In this way, a copy of the certificate of authenticity can be used to pass off as a genuine a non-original work.
Figure 1 shows some examples of "classic" paper certificates of authenticity, and it is pretty clear to see they are relatively easy to fake.Generally, (

Introduction to Hylemetry
The traditional ways for establishing the authenticity of sensible objects, such as documents, banknotes, packaging and high value products rely on the presence of secret identifiers or on complex's manufacture process, which is hard to overcome, or counterfeit.Classic examples are barcodes, holograms, RFID, etc.
In any case has to be considered that the difficulty to duplicate, it does not mean inability to duplicate: "what one man can make, another can copy" (Nmab, 1993).Observe that conventional RFIDs could be easily copied and counterfeited and thus, are not capable of resolving the problem or tag authenticity (DeJean & Kirovski, 2006, 2010).
Currently, the unique identification of persons process is based on use of non-reproducible physiological or behavioral characteristics.Distinctive biometric features enclose fingerprints, hand geometry, retina, iris patterns, voice waves, DNA, handwrite signatures, etc.; this methodology is called biometric authentication (Woodward et al., 2002).
Biometric authentication is centered on the acquisition of individual human physical qualities, used to compose a template (feature vector) for further biometric identification (Woodward et al., 2002;Jain et al., 2011).This template is compared with that extracted during the testing phase to determine whether a person is who he claims to be.Because of difficulty of acquiring in different phases, and with different acquisition systems, the same model, a verification method based on a set of thresholds is expected.
For an inanimate object, any model made to a particular technology can be replicate employing like tools.On the other hand, each pattern generated by random processes is a feature not replicable that can be exploited for hylemetric identification (Melen, 1999;Zhu et al., 2003).

Hylemetric Characteristics
In biometric authentication, the acquired feature should have the following properties (Cozzella et al., 2012b):  Universality: characteristic has to be present in all the individuals;  Permanence: characteristic stable over time;  Uniqueness: samples related to different individuals should be as unlike as possible;  Robustness: samples related to the same individual should be as close as possible;  Availability: easy to acquire with a dedicated sensor;  Acceptability: people shall consider the acquisition method as nonintrusive;  Forge Resistance: the acquisition system should be arduous to deceive.
A system used for authentication of non-living objects (hylemetric authentication), should have similar properties.
In general, in hylemetric verification the feature vector (template) should have the following characteristics (Clarkson et al., 2009):  Exclusivity: Each object must be distinguishable from all others;  Stability: feature vector should be confirmable by multiple entities for all lifetime of the object;  Shortness: template should be short and easily computable;  Resistance to imitation: feature vector should be hard or impossible to duplicate; an object in such a way that the clone to express the same template as the original one.
Each random or hard-to-reproduce texture, can be potentially exploited in as hylemetric features.Good hylemetric features have to content also the following necessities:  it has to be simple repeatable and reliable to extraction the feature vector;  the template must be manufactured at low cost, compared with a chosen level of security;  the cost of exact or near-exact replication of the random structure employed as hylemetric features has to be significantly greater than the commercial value of the cloned object;  the cost of the authenticity verifying procedure has to be small, again compared with a chosen level of security. www.ccsen

Hyleme
In any artw artworks, process.A structures the same i hylemetric random pa function (G from the r from three If we refe acquisition geometrica  In this work, we have used affine transformations to achieve the correction of geometric distortions.Verification software, after having acquired the trust points, applies the correct transformation, which allows obtaining a corrected version of I T to be used during matching phase. After having applied geometrical transformation, it is possible extracting the HHP T from the transformed image I' T .The process is the same used for creating the HHP C related to certificate image (the necessary procedure to obtain HHP T must be reported on the digital certificate authentication).
The last step consisting in matching HHP C and HHP T , the Hylemetric Hash Pattern extracted from the digital authenticity certificate (and related to I C ), and from the geometrically corrected test image I' T respectively.Due to possible residual geometrical distortion and presence of noise, it is usual to obtain a HHP different in comparison to that present in the authenticity certificate, also in case of original artwork verification.Therefore, considering that the HHP has a casual structure, to allow the comparison among HHP T and HHP C , in this paper a verification approach based on digital cross-covariance calculation is proposed, similar to the one used in speckle field measurement (Sjӧdahl, 2000).
In this work, the used cross-covariance formula is: In Equation (1) (∆x, ∆y) represent the correlation peak coordinates, where F and 1 F  are forward and backward Fourier Transform operators, respectively, and * indicates the complex conjugate.The Equation ( 1) can be efficiently calculated using a Fast Fourier Algorithm (Brigham, 1988).The coefficient α has the aim to control the correlation peak width.Optimum values range are from α = 0 for image characterized by high spatial frequency content and high noise level, to α = 0.5 for low noise image with less fine structure.For α values greater than 0.5 the high frequency noise is magnified.In our experiment we have always used α = 0.5 values.
As in biometric approach, also Hylemetry introduces a correlation threshold, necessary to define if the two HHP are similar enough to be considered the same.The threshold used in this paper is defined as follow: false artwork .genuine artwork The appropriate threshold is chosen in such a way as to minimize the False Acceptance Rate (Jain et al., 2011), such as the percentage of fake artworks identified as genuine, respect the total quantity of authentication checks.It has to be observed that the introduction of geometrical adjustment has highly reduced False Rejection Ratio, due to genuine artwork recognized as counterfeited.
Figure 4 shows the previously described process in a visual and schematic way.

Conclusion
In this paper an innovative methodology, founded on Hylemetry, was demonstrated, which allows to produce not duplicable certificate of originality.The chosen hylemetric characteristic allows to obtain high verification rates, thanks also to a geometrical correction preprocessing, executed in manual way, to correct errors and misalignments of acquisition during the verification procedure.The proposal method allows to generate a unique, Hylemetric Hash Pattern, used for verifying object authenticity.The usage of a well-known and highly consolidated speckle metrology matching algorithm allows to be confident on the obtained results and the robustness of the procedure.It is also possible to implement a securing method, based on DSS, for protecting digital certification of authenticity from copy attack, the most used method in this situation.Future studies for obtaining an automatic system able also to determine an optimum threshold, are in progress, with the aim to be independent from the interested object.
Very recently, the van Gogh museum and the FUJIFILM Belgium have developed a new and unique reproduction process able to reproduce the textured detail of Van Gogh's artworks in accurate color.This new method of reproduction of the paintings allows you to achieve the best fine art replicas ever seen.These copies replicate exactly the format, the color brightness, and texture of the original (Fujifilm, 2013;Van Gogh Museum, 2013).In future studies we want to verify whether our method works well on these copies of the highest quality.
In particular we will try to check if you can distinguish a copy from the original and a copy from another.

Figure
Figure 3(a (Figure 3( verificatio The next s Figure 5 r (I C (LD) ); thi identify th from the o original on the reporte the image finally rep (Figure 5(f In Figure geometrica = 0.31, the have a nor the previou