We live in a society marked by the digital transformation in which more and more procedures are being transferred to an online environment. From a loan application to renting a vehicle or betting on our favourite team, the options, facilities and advantages that the digital world offers us are endless but… are they safe?
The number of cases of identity theft and document forgery increases over the years. In fact, the latest researches reveal that complaints of identity theft in Spain are increasing at an average annual rate of 16.5%, causing an unsustainable economic and reputational risk for most companies in the country. Thus, reducing the risk and preventing fraud becomes the companies’ greatest concern, and technology advances in multimedia security are their best allies.
Daniel González (DG), Diego Pérez (DP) and Pablo Dago (PD) talk to us about the importance of multimedia forensics, its evolution, use cases and sectors in which it has already been implemented. In this issue of GRD Tech, they reveal the most effective tools for the prevention of document forgery.
What is multimedia forensics and why is it important in 2020?
PD: Multimedia forensics consists of analysing all types of media content to find out the type of processing or operations that have been done on them. Why is it important? Mainly because everything is moving to the online world. The operations that used to be done in physical space, going to an office and presenting the documentation, are now being done over the internet, and that makes fraud much easier to hide.
Furthermore, in situations such as the current health crisis in which the migration of many businesses to the online environment has been accelerated, this technology is more important than ever, as the more information an online company receives from its customers, the higher the risk of a fraud being committed.
DP: Currently, the generation of deepfakes, which allows false videos to be artificially created, is also having a great impact by, having a person do or say things that they have never really done or said. These techniques, although they have many beneficial uses for example in the film industry, are also being used for fraudulent purposes. In our case, within the framework of EGIDA (the National Network of Excellence in Security and Privacy Technologies that we lead), we are working on an algorithm capable of detecting this type of videos.
What are the uses of this technology? Which sectors can benefit from its implementation?
DP: Originally, the use of multimedia forensics focused primarily on analysing and revealing fake news. Nowadays, with the digitalization of all processes, especially in the financial sector, insurance companies, and online gambling houses, it is increasingly common to find modifications in documents, especially in the images that are part of the digital onboarding process such as ID cards. Solutions based on multimedia forensic technologies allow to tackle this problem by detecting the forged of documents or images.
DG: These multimedia security technologies have application in everything related to the validation of digital procedures related to administration, invoice payment, etc. In short, in any procedure that requires sending documents in a digital format.
PD: Due to the changes that have been made in its legislation, the financial sector is one of the main users of this technology. In the next few years it is very likely that shared mobility will become one of the main sectors of application for the validation of identity documents and driving licenses.
What is Gradiant working on right now?
PD: We are working on the detection of fraud in images, documents and videos. In the case of images, we detect manipulations in scanned documents and in any kind of photographs, such as ID or passport pictures.
DP: Furthermore, although initially we aimed to digital manipulations, for which we already have an effective detection system, we are now also focusing on physical manipulations, which applies to identity and text documents in general. Right now we are investigating the use of algorithms for detecting font inconsistencies, something that can be detected despite the fact that the modification has been made in the physical format, both at the PDF and image level.
How did Gradiant get here?
DP: At a documental security level, we have worked on several projects that allow us to guarantee document integrity and traceability. One example is SIGNED project, an European project whose objective was to bring the digital signature to the printed domain of documents; or Shadow, a tool for document traceability in printed documents that we sold to Telefónica in 2016.
If we focus on the technologies more directly related to forensics, we can highlight FORENSICA, included within the TACTICA Project, in which we developed a library that implemented different methods for forensic analysis of images and videos. In addition, we have accumulated years of experience in the use of artificial intelligence techniques such as machine learning and deep learning that apply also to forensic analysis algorithms.
DG: This is also an area in which there has been special collaboration with the GPSC group of the University of Vigo.
The first digital forensic analysis methods appeared at the end of the 70s. How has the technology evolved so far?
DP: Until now it was a very academic field with low TRLs, and it did not really reach the market. There were many publications but it was very difficult to generalize those methods and use them in real environments. During the last few years, with the increase in the use of deep learning, it is beginning to make the jump to the market and the real world.
DG: That’s right. There were many academic works but with very restricted databases, which meant that they didn’t correspond to the real world. On the other hand, the evolution of the technology and the digitalization process also play an important role. There is more and more fraud in the digital environment, which makes these techniques more necessary than ever.
What trends or new uses will we see next year? What challenges will this technology face?
DP: On a global level, what has the most impact right now is the detection of deepfakes. In fact, Facebook, along with Microsoft, Amazon and several universities, launched a contest with a $10 million prize to promote research into deepfake detection, although the results obtained have not been very relevant. This is a very prominent problem in which many companies with open research teams are looking for a solution. It is a complicated issue, since at the same time, new and better deepfake generation algorithms are being developed to avoid detection.
DG: On a regional level, the main challenge is the document validation that we spoke about earlier. The detection of physical manipulations is a field in which it is necessary to improve the technology, and it will continue to play a fundamental role. It will therefore be a trend, and it will also be a challenge. It is a field in which there is a constant need for research, since there are always new ways of manipulating certain documents, new tools for generating deepfakes, etc.