Digital imaging: Insurers, publishers, law-enforcement agencies and dating sites are using software that can detect the digital manipulation of photos.
The photo splashed across front pages worldwide in July 2008 showed four Iranian test
missiles blasting skywards. 
Released by the media arm of Iran’s Revolutionary Guard, Sepah News, the picture  was soon found to have been manipulated: 
One missile had been cloned and appeared twice, evidently to conceal the fact that another had failed to lift off. 
Governments have long doctored photos for political reasons. 
In Nazi Germany and Communist Russia, senior figures who fell from favour were commonly airbrushed out of photographs. 
Now, thanks to digital technology, image manipulation is available to everyone, and nefarious uses are becoming far more widespread.
In around one in 75 insurance claims, photos documenting property damage have been fraudulently retouched, says Eugene Nealon of Nealon Affinity Partners, a company based in London that advises insurers. 
Liz Williams, editor of the Journal of Cell Biology, says her publication rejects around 1% of peer-reviewed scientific papers after discovering that microscope images have been doctored to make results look good.
Efforts to automate the detection of doctored images are bearing fruit. 
Last year Fourandsix Technologies, a start-up based in Silicon Valley, began selling an add-on for Photoshop, called FourMatch, that determines whether an image has come straight from a camera or has been manipulated. 

It compares the “metadata” associated with the image against a database of signatures that represent the characteristic ways in which different devices capture and compress image data, to ensure that the image is what it claims to be. 
FourMatch is sold primarily to law-enforcement agencies and costs $890. 
But it cannot tell whether a manipulated image has been slightly tweaked or extensively doctored. 
So a human analyst is still needed “in the loop”, says Mr Farid, one of the firm’s co-founders.
A trained eye can spot inconsistencies in shadows, reflections and incorrect perspective.
Many fakes are obvious. 
But unmasking a sophisticated forgery can require hiring an expert for days, says Hany Farid, professor of image forensics at Dartmouth College in New Hampshire, who sometimes acts as an expert witness. 
Rockefeller University Press, publisher of the Journal of Cell Biology and other journals, employs a full-time analyst to examine submitted images, opening them in Adobe Photoshop and looking for anomalies by adjusting a series of display settings.
Another tool, recommended to insurers by Mr Nealon’s firm, is Verifeyed, made by a company of the same name based in Prague. 
Like FourMatch, its software examines image files to look for inconsistencies in colour, encoding and compression formats that indicate the use of retouching software to modify the image. 
The software costs $899, and Verifeyed will also analyse large batches of photos on behalf of clients, charging around $0.07 per image. 
Its customers include insurers, law-enforcement agencies, forensics labs, a dating website and a logistics company keen to expose bogus claims that goods were damaged in transit.
Verifeyed’s boss, Babak Mahdian, reckons that the technology could also be useful to news organisations that wish to check that photographs are genuine.

Manipulation-detection software is desperately needed in health insurance, says Mr Nealon. 
In some countries 2-3% of claims submitted contain images retouched to “embellish the truth”, he says, making injuries look worse than they really are in order to collect more money. Only about 10% of insurers are using software to spot fakery, says Mr Nealon, so there is plenty of room for growth.
In January a new law came into force in Israel making it illegal to use images in advertisements that have been retouched to make models look thinner without printing a disclosure on the picture. 
Online images are exempt, however. “We’re not going to correct the whole world,” says Rachel Adato, an Israeli parliamentarian who sponsored the bill. 
The “Photoshop law”, as it is known, has prompted efforts to pass similar legislation in America.
A feature introduced several years ago by Canon and Nikon, the two leading camera manufacturers, gives photographers a way to prove, if challenged, that their images have not been manipulated. 
When a picture is taken, the cameras attach a coded signature that is destroyed if the image is modified and resaved. 
An intact signature, then, should prove that a photo is genuine. 
But researchers at ElcomSoft, a computer-security firm based in Moscow, have shown that the system is easily fooled. 
Counterfeiters can copy an image’s security signature and reapply it after retouching, says Vladimir Katalov, ElcomSoft’s boss.
Another way to determine whether an image has been manipulated or not relies on the fact that in a digital camera’s grid of millions of light sensors, several are usually flawed. 
Each flaw creates a tiny discolouration, imperceptible to the naked eye, in pictures taken by the camera. 
If the pattern of unusual discolourations in two pictures taken by the same camera matches exactly, neither image has been retouched, at least in those areas. 
But the analysis is complex and time-consuming, says Anderson Rocha of the University of Campinas, near São Paulo, who provides photo-forensics advice to law-enforcement agencies in Brazil. 
His team is developing software to automate the mapping and matching of discolouration patterns.
Software can generally sniff out amateur retouchers, says Dr Rocha, but professional forgeries continue to slip through. 
This is partly because skilled forgers keep up with the academic literature on image forensics, says Siwei Lyu, a researcher at the University at Albany. 
Forgers now have the upper hand, he says.
The ultimate solution may be to separate photographers from their photographs, so that they do not have the opportunity to manipulate them, says James DeBello of Mitek Systems, based in San Diego. Bank of America, JPMorgan Chase, Wells Fargo and other big American banks let customers deposit cheques remotely by taking pictures of them with their smartphones. 
They do so using an app provided by Mitek that takes a snapshot of the cheque and sends it straight to the bank, leaving nothing on the phone itself.
Codasystem, a French start-up, offers a similar service. 
Its “Shoot and Proof” system allows an image to be taken with a mobile device, tagged with a timestamp and location and then uploaded to an online repository. 
The image can then be viewed and downloaded, but the original remains in the repository, untouched. 
Any nefarious Photoshoppery can therefore be easily exposed. 
Digital technology can, it seems, be used to prove that an image is genuine—even as it makes it easier than ever for cameras to lie.