Within the digital forensics realm, it’s not uncommon for us to be asked whether or not we can determine if a photograph or image is fake or has been altered from its original state. Typically, the first place we look to analyze is the internal metadata of the file to determine if any of the values have been added, altered or been updated. Faking images is only getting easier to do, especially by the average user.
So, how do we combat this problem? Researchers at New York University’s Tandon School of Engineering recently authored a study – “Neural Imaging Pipelines – the Scorge or Hope of Forensics?” The basis of the study is that machine learning can be used to spot fake photos and suggests the detection method be included right into the source – the camera.
Essentially, every photo taken has a digital watermark to determine its authenticity, akin to a chain of custody, that can be used to determine whether the content is authentic. There is a lot of advanced technology in this study, including the usage of “neural imaging pipeline” that learns to include artifacts onto high-fidelity images during the processing of the image.
There are a couple of potential roadblocks to replacing well-established digital imaging technologies, and the foremost is whether camera manufacturers will employ photo processes using machine learning and the “neural imaging pipelines” needed to digitally watermark photographs during the imaging process.
There is a clear need to develop better processes to determine the authenticity of an image or photograph, but you can be sure the bad guys will figure out a way to circumvent the new technology.
Email: Phone: 703.359.0700
Digital Forensics/Cybersecurity/Information Technology