Until now, scientific analysis has been an expensive, time consuming and often inconclusive option for galleries, collectors and advisors. Our tests, designed and executed in collaboration with leading laboratories, address these shortcomings. Carefully sequenced tests eliminate the possibility that the work was produced by an enterprising forger, and further tests confirm the work's attribution. Sophisticated techniques with the greatest efficacy are prioritised, meaning unnecessary tests are avoided.

Unlike other companies that prioritise inconclusive pigment characterisation and chronologies, Hephaestus specialises in employing well-defined dating techniques to support attributions and detect forgeries. A single microgram sample of paint is taken, together with a strand of canvas: the material is analysed and our proprietary statistical modelling provides an accurate date for the work.* If consistent with the proposed date of the artwork, a strong case for attribution can be built.
‍
Eliminating the potential for forgery and placing the work within a particular time period means that we protect the judgement of connoisseurs, advisors, galleries and collectors. Further analysis can then be prescribed to add to the body of evidence that a painting is by a particular artist. Further tests might include: 

• Imaging (x-radiography, IR, UV)
  
• Pigment Analysis
• Identification of the binder

*N.B. a forger sourcing and using old paint, would not be able to fool this test.

Machine Learning: the future of authentication

A kind of “arms race” has emerged, pitting art emulation against detection. Authenticators are falling far behind forgers. The number of connoisseurs is diminishing, and detection technologies are largely unchanged from where they were two decades ago.

In its determination to win the technological “arms race”, Hephaestus is also developing the next generation of tools to detect misattributions and eradicate forgery. In so doing, Hephaestus has turned to the field of astronomy, where machine learning (artificial intelligence) is combined with more traditional tools of observation to recognise patterns – of light diffusion, of chemical composition and gravitational pressures – undetectable with our five senses.

One of the best-known accomplishments of machine learning is quantifying structures or objects with complex features. Adapting these technologies for more terrestrial concerns, Hephaestus has developed technological tools that identify a painting, or a portion thereof, as the product of a singular artist, and eliminate those works which do not belong to the artist’s oeuvre. To give one example, our naked eyes cannot detect the characteristic patterns with which a painter applies paints across a canvas, but a machine can. When machine learning pattern recognition combines with chemical analysis, provenance review and traditional connoisseurship, we reach the highest evidentiary standards available. While no method of authentication can be said to be fool-proof, forgers facing Hephaestus are duelling with an arsenal of connoisseurship, laboratory testing and the statistical methods from the very frontiers of science. 

The suite of Hephaestus’ tools offers important advantages over conventional methods for collectors, museums, lenders and insurers:

• We are able to test rapidly.
  
• We are able to test more accurately and significantly more affordably than alternatives. 
• We are able to test on-site, with minimal or no invasion of the artwork itself, lowering transportation risks and increasing savings.