Manipulating photographs has been a feature of photography ever since it was first invented 200 years ago.
Unfortunately, it is practically impossible to detect a fake or photoshopped image with 100% accuracy.
The software editing tools and techniques currently available are simply too sophisticated and powerful, and prevent image forensics tools from achieving such a high level of certainty.
However, there are a number of tools and techniques you can use to reach a 60-80% certainty that an image was modified.
The most common way to detect photoshopped or modified images is to use specialized photo forensics tools that detect and highlight tiny differences between pixels, often caused by photo editing. Other methods are analyzing EXIF data or training yourself to find evidence of photo manipulation.
7 Tools to detect photoshopped images
1. Forensically Beta | Browser | Free
Link: Forensically Beta
Forensically Beta is perhaps the best photo Forensics tool available right now, and best of all it’s completely free.
Forensically has multiple image detection tools that can help you figure out if an image was modified or not.
The Error Level Analysis is the most important tool, and most commonly used in photo forensics to identify modified images.
ELA scans an image to look for differences in how pixels are compressed. In non-modified images, all pixels are at nearly identical levels of compression.
However, photos that are modified have patches of pixels that are at a different compression level. As a result, they stand out during Error Level Analysis, like in the image below.
Other tools included in Forensically are:
- Clone Detection. It identifies subtle repetitive patterns in a photo that possibly indicate tampering.
- Noise Analysis. Removes most of the image and leaves only the “noise”, which can highlight brushing, warping and other deformations.
- Level Sweep. This will change the contrast of the photo to make it easier to find edges of elements that were photoshopped into the image.
Forensically Beta is overall quite complex, so its best to check the tutorial whenever you are using it.
Keep in mind however that this tool isn’t a magic bullet. It won’t automatically reveal if an image was modified or not. You will still have to interpret the results yourself and come to your own conclusion.
While here is the same photo quickly modified in Paint, alongside it’s Error Level Analysis.
2. Ghiro | Desktop Program | Free
Ghiro is a free image forensics tool designed to analyze thousands of images at once and pass them through a series of filters designed to identify manipulated images, such Error Level Analysis or Metadata Analysis.
Other interesting tools are:
- Hash Matching. Basically, every photo is identified by a “hash code”. Ghiro can scan the Internet for this hash code to find if the image was already identified as fake by others.
- GPS localization & Metadata. In some cases, Ghiro can even tell you where and when a picture was taken, or even what device was used.
Ghiro is easy to use, but its installation process is a bit wonky so we suggest the following guides:
3. ImageJ | Desktop Program | Free
ImageJ is an image analysis tool developed by the United States National Institutes of Health that is designed for scientific accuracy and uncovering photo manipulations.
It supports operations like edge detection, contrast manipulation, the measuring of distances and angles, sharpening, smoothing, rotation and many other and it allows the processing of stacks.
Below is an example of an image that was modified, and how ImageJ was able to spot the modifications.
We then processed the image in ImageJ, which automatically identified and annotated key areas of the photo.
Finally, ImageJ, using custom settings, outputs an analysis that highlights how some of the elements on the image are identical.
For example, it found that elements 1 and 4 as well ass 7 and 11 are identical.
Unfortunately, ImageJ is rather complicated to use and requires you to have at least a basic-to-intermediate understanding of digital photography.
Fortunately, there are some resources to help you learn how to use and interpret ImageJ, at least as a starting point:
Overall, ImageJ is quite technical and difficult to use, so learning to use it is a skill all of its own.
4. Foto Forensics | Browser | Free
Link: Foto Forensics
Foto Forensics is similar to Forensically, but its only fake image detection algorithm is Error Level Analysis, while Forensically has multiple .
Still, it’s a completely free tool and it works great if you want to cross reference with Forensically.
5. FakeImageDetector.Com | Browser | Free
This site is an AI powered tool that tries to detect fake, modified or AI generated images.
In my experience, the accuracy is so-so:
That being said, the site can help in detecting photos that are made by AI image generators or other computer generated images.
6. JPEGSnoop | Desktop | Free
JPEGSnoop is a free Windows program designed to perform image forensics, which makes it similar to Forensically and Foto Forensics.
Besides image forensics, JPEGSnoop can also be used to assess the quality of images, locate errors, determine what editing program was used to alter the photo, etc.
What makes JPEGSnoop more useful than either Forensically Beta and Foto Forensics is that JPEGSnoop divides images into four classes, where Class 1 means a photo was 100% modified, while Class 4 means a photo is “probably” original.
Overall, if you’re a technical person and know how to read detailed photo information, then JPEGSnoop is probably the best free forensics tool.
7. Amped Authenticate | Desktop | Paid
Link: Amped Authenticate
Amped Authenticate is part of a forensics software suite that is specifically designed to identify modified images.
Besides ELA, Amped Authenticate also uses other methods to identify fake images, such as:
- Geometrical Analysis of lights, shadows and perspective which can reveal an image is possible or not.
- Camera Identification. Every camera produces a unique and distinctive type of noise, like a fingerprint, and this program can identify the camera based on this unique noise pattern.
- Algorithmic Analysis. Through its own proprietary algorithms, it can identify suspicious pixel patterns that suggest a photo was manipulated.
Amped Authenticate is a paid program, and an expensive one at that, so it is more likely limited to corporate customers.
8. Online EXIF Viewer | Browser | Free
Link: Online Exif Viewer
Every single image on the Internet has a set of information parameters attached to it called EXIF data.
Certain EXIF data parameters can help you figure out if an image is genuine, modified or completely fake, such as:
- When the photo was taken.
- When the photo was modified.
- What device took the photo.
For example, the EXIF data below tells you an image was modified because the DateTime and DateTimeOriginal are different, meaning the photo was at some point modified.
- Software: the specific photo editing program that modified the photo.
- DateTime: the time when the final version of the photo was made.
- DateTimeOriginal / DateTimeDigitized: the time when the original non-modified photo was created.
Unfortunately, some images don’t have useful EXIF data attached.
This is because there are a number of ways you can remove an image’s EXIF data such as:
- On PC, you can simply remove the EXIF data by right click image -> Properties -> Details -> Remove Properties and Personal Information.
- Screenshotting the photo.
- Uploading it to various sites.
In this case, the best approach is to just assume that if a photo doesn’t have EXIF data, then it was probably modified or altered in some way.
9. Use Google Reverse Image Search | Browser | Free
Reverse image search engines such as Google or TinEye can help you see if an image has previously appeared on the Internet.
By using reverse image search you can figure out the context of the image, and whether other people say it is real or not.
A good example of this is the famous AI generated image of Pope Francis in a puffy jacket:
The downside to reverse image search is that not all images are indexed in search engines.
For example, images on Facebook, Whatsapp, Telegram, Discord etc., don’t appear on search engines.
In this case, you’ll have to rely on other methods, such as pixel peeping, EXIF data, image forensics software etc.
10. Hone your fake photo detection skills
Like most things, detecting fake and modified images is a skill. There are even websites that make you guess whether an image is fake or not.
The point is that even without sophisticated photo forensic tools, there are several things you can look at to determine if an image has been modified or not:
- Be aware of human anatomy: Sometimes, a photo doesn’t even need to be photoshopped. Instead, the person just needs to know how to use human anatomy and angles to dramatically change their appearance.
- Inconsistent Lighting and Shadows: The direction, length, and color of the shadows in an image should be consistent with the light source(s). If they are not, it’s possible that the image was manipulated.
- Edges and Borders: Edges of objects in the image should appear smooth and natural. If they are overly sharp or blurry, it might be a sign that the image has been manipulated.
- Texture and Noise: In an unedited image, the texture and “noise” (graininess) should be consistent throughout the image. If certain areas look different in texture or noise, they may have been added or altered.
- Perspective: The perspective of objects in the image should be consistent. If an object appears to be at an odd angle or seems out of place, it could be a sign of manipulation.
- Repeated Patterns: Many editing tools use a cloning technique to cover up manipulations. If there are repeated patterns in the image, it could be a sign of cloning.
- Image Quality: If parts of the image have different resolutions or levels of clarity, it could indicate that those parts were added from another image.
- Unnatural Colors or Lighting: If the colors or lighting in a part of the image seem unnatural or don’t match the rest of the scene, it could be a sign that the image has been manipulated.
- Too Perfect: If a photo looks too perfect, it might have been edited. For instance, in a group photo, it’s unusual for everyone to have their eyes open and be smiling perfectly.
That being said, while these signs can suggest an image might have been manipulated, they are not definitive proof.
This is because highly skilled photo manipulators know how to avoid these issues, so their photos may not exhibit these signs at all.
However, these checks can still be helpful when you’re analysing an image’s authenticity.
- 3 Ways to Forward Emails to your Phone as SMS Text Messages - December 15, 2023
- 5 Ways Gmail Prints Money As a Free App - December 15, 2023
- Is A 100 or 200 Mbps Connection Fast Enough for Gaming? - December 15, 2023