Regular readers of this blog have learned about a diverse collection of image forensics techniques which Hany Farid and others have developed over the past decade or more. Yet Fourandsix currently offers just one commercial product featuring just one of these many techniques. Of course, this type of situation is common in technology. Any followers of tech blogs have no doubt seen countless articles featuring the latest innovations in research, only to discover that they’ll need to wait years to access those innovations in a product, if indeed the ideas ever get productized at all. Why is it such a bumpy road from research to final product, and what, you might ask, is Fourandsix doing to bring more technology to market?
Fundamentally, the difference between research and product development is that the former merely determines whether something is possible, while the latter requires figuring out how to make it practical. Research typically starts with a general problem and an idea of how to solve it. The research is, in some sense, “finished” once the researcher can demonstrate that the idea will work.
Getting from research to product, however, requires answering a slew of additional questions and considering a variety of complex issues. In the case of software products, here is just a small sampling of the types of questions that need to be asked:
Without the right answers to these and other questions, the research won’t actually be practical. If users become too easily frustrated with the product or service—because it’s too slow, too confusing, or too erratic in its operation—then the product will fail.
This is as it should be. If researchers worried about all of these questions at the outset, they’d likely halt many ideas before they were fully explored. When freed of these concerns, they can more effectively point the way to what’s possible. It’s then left up to the product people to figure out if there’s a way to make some of those possibilities practical for everyday use.
With our first product FourMatch we focused in relatively narrowly. We took a single image forensics technique—JPEG signatures—and created a product for specialists to use in determining whether a file is an unmodified original from a digital camera.
For much of the past year, however, we’ve been focused on what’s next. How can we start incorporating more forensics techniques and package them up in a way that will be useful to more people? The gap between possible and practical is still a bit too wide for a few of the technologies that I personally think are the coolest, though that gap may well shrink over time. In the meantime, though, we’re taking some of the more immediately practical technologies and doing something new with them. Stay tuned…