Bayesi-Chain: Intelligent Identity Authentication

In a bid to stamp out fraudulent crime, there is increased pressure on individuals to provide evidence that they possess a ‘real’ identity. Counterfeiting and fake identities have reduced confidence in traditional... [ view full abstract ]

In a bid to stamp out fraudulent crime, there is increased pressure on individuals to provide evidence that they possess a ‘real’ identity. Counterfeiting and fake identities have reduced confidence in traditional paper documentation as proof of identity and has created a demand for an intelligent digital alternative.  Recent government implementations and identity trends have also improved the popularity of digital forms of identification.

Authentication of identity is a salient issue in the current climate where identity theft through record duplication is on the rise.  Identity resolution techniques have proven effective in filtering duplicated and fake records in identity management systems. These techniques have been further improved by the implementation of machine learning techniques which are capable of revealing patterns and links that have formerly gone undetected.  Research has also suggested that incorporating non-standard attributes in the form of social contextual data can increase the efficiency and success-rate of these fraud detection methods.

In the digital age where individuals are creating large digital footprints, online accounts and activities can prove to be a valuable source of information that may contribute to ‘proof’ that an asserted identity is genuine.  Online social contextual data – or ‘Digital identities’ -- pertaining to real people are built over time and bolstered by associated accounts, relationships and attributes. This data is difficult to fake and therefore may have the capacity to provide proof of a ‘real’ identity.

This paper outlines the design and initial development of a solution that utilizes data sourced from an individual’s digital footprint to assess the likelihood that it pertains to a ‘real’ identity.  This is achieved through application of machine learning and Bayesian probabilistic modelling techniques.  Where identity sources are considered reliable, a secure and intelligent digital identification artefact will be created.  This artefact will emulate a blockchain-inspired ledger and may subsequently be used to prove identity in place of traditional paper documentation.