Google just figured out a way to show how hackers might be able cheat stock photos from their watermarks. The company has also managed to figure out how to stop the piracy from happening with a quick fix as well. Stock photos, despite being well-equipped for any and every occasion, come with hefty subscription fees that small agencies and individuals may not be able to pay for, when they create their collateral. Here is how Google is doing it.
The search giant has developed an algorithm that can remove the watermark from stock photos, all in a matter of seconds. All of this can be done automatically as well and requires no extra effort form the user. Researchers from Google have been able to pin-point exactly how the watermark is attached to the image in the first place. The algorithm takes into the account of the flaws that the producers of the stock photos make when they add the watermark.
Apparently, Google has figured out that the watermarks are created by default and all feature the same pattern and properties that are copied and pasted onto each image at a time. This way, hackers will just need to know the relative properties of each image and then undo the watermark effect to restore the image’s original charm. The only way that stock photo platforms will be able to counter this method, is by making their watermarks more complicated than before, such as adding warps to their watermarks or their images. All of these subtle changes to each individual image can make it nearly impossible for hackers to cheat the stock images. Watermarks purposely already feature lines and shadows that make it difficult to duplicate.
The randomization of these watermarks and the effects added, may or may not be surpassed in the near future, as there could be newer ways discovered that would be able to bypass the security that the watermark offers. Google being the good Samaritan it is, has already published its findings and the techniques involved that hackers may use in order to refrain from paying for stock images and have posted a blog and also published a paper on their research. The research work was conducted by Tali Dekel, Michael Rubenstein, Ce Liu and William T. Freeman. They used AI to teach itself how to understand the structure of the watermark in the image and how to counter its effects.