Entering the Digital Customer Onboarding Era: How the Semantic Web Can Help

Abstract

In this presentation we showcase the use of Semantic Web in AriadNEXT's IDCHECK.IO document veri cation service. This service has been introduced for a number of years now. It has recently seen a speed up in its market adoption due in large amount to the introduction of a new semantic web data model work ow. We will start by introducing the research project behind this technology upgrade, then explain our approach, focusing on what problems the use of semantic web solves, and nally give some highlights of the perceived business bene ts.

Introduction

The IDFRAud [3] project is an industrial research project led by French ID document veri cation leader company AriadNEXT, that provided the technology behind its document veri cation service IDCHECK.IO [ 2 ]. One of the objectives of IDFRAud is to propose an automatic solution for ID veri cation that can handle documents issued from a large set of countries. The solution should be able to execute speci c controls according to the ID model (type, country, generation, etc) thanks to a knowledge base. The core idea of IDFRAud project is to provide an automatic veri cation system for identity documents in order to replace existing manual veri cation processes. The di erent components of ID analysis and veri cation in IDFRAud are driven by a set of control rules. In order to guarantee an interpretable and adaptive behavior at each ID analysis step, the identity document descriptions are organized by a knowledge management module.

Technical approach

One of the requirements of our knowledge management module is its interoperability and portability. It is preferred to store the data in a standard way in order to be able to use other tools such as Sewelis [4] to navigate our data. We also have another strong requirement: we must be able to easily extract a subset of the knowledge base to run on mobile platforms, where C/C++ language rules. After a thorough analysis of existing technologies, we decided to use RDF for its versatility/ exibility of knowledge modelling using graphs, coupled with its capability to bring formal structure to the knowledge using Web Ontology Language (OWL). RDF serialization formats are also strongly standardized, as a di erence to many database systems. This standardization encouraged engineers to write RDF processing tools in many di erents langages, including those available on mobile platforms.

Providing an e cient, consistent, and descriptive enough model for ID documents proves to be a very challenging task. The biggest issue encountered is the very high diversity of how ID documents look like, which makes it really di cult to design a data model that ts all cases. As such it is anticipated that the model will see serious evolutions with the number of supported documents.

We decided to build two in-house frameworks to ful ll those requirements: 1. ModelEditor is a UI that provides guided data editing. It enables a non-expert user to ll in data that is consistent with the underlying OWL ontology. It can be made fully generic, as the UI is built dynamically for the OWL ontology. 2. AutoRDF [ 1 ] is a code generation utility which makes code base maintenance easy by being able to follow a constantly evolving ontology. AutoRDF can be of some use to any kind of project that needs to manipulate RDF data that conforms to an existing ontology. The use of modern C++ makes it portable to a wide variety of platforms, including all mobile phone platforms, as well as most of the embedded world systems.

The steps used in our document de nition process are now the use of Protege open-source ontology editor for formal concepts de nition, the use of ModelEditor to input new ID model data, and lastly AutoRDF [ 1 ] to code processing logic using the newly de ned concepts. We have experimented that those three steps make for an e cient document description environment. 3

Business bene ts

We will give a few direct an indirect metrics of the performance gains of this approach. We will give some key gures: { Decrease in time needed to specify and support a new document model, from 6+ month to only a few days. { Increase in business turnaround, including an astonishing 800% year to year peak growth rate. 3. IDFRAud: An Operational Automatic Framework for Identity Document Fraud Detection and Pro ling - Joint research project with AriadNEXT, IRISA, ENSP and IRCGN funded by ANR grant ANR-14-CE28-0012. http://idfraud.fr/ 4. Ferre, S., Hermann, A.: Reconciling faceted search and query languages for the Semantic Web. Int. J. Metadata, Semantics and Ontologies 7(1), 37{54 (2012)

References

1. AutoRDF - A framework for C + + proxy class generation from Web Ontology Language . https://github.com/ariadnext/AutoRDF

2. IDCHECK. IO - Fast and reliable ID document checking . https://idcheck.io