Error Level Analysis can detect digital image modification

Error Level Analysis can detect digital image modification is about image analysis with Error Level Analysis (ELA). ELA is an online tool found on fotoforensics.com. It is easy to use and a relatively useful tool to detect digital image modification. Analysis of images with ELA is free and can be used for image formats like JPEG, PNG, and WebP.

Error Level Analysis can detect digital image modification

The image above shows five “fake” helicopters inserted into a photo using Photoshop CS4. Immediately you can not see the deception but with ELA (Error Level Analysis), you can examine the photo in depth and reveal the scam.

Image Analysis with ELA (Error Level Analysis)

Analysts at all levels can use ELA for image analysis. The program is extensively used in UFO circles to detect fake UFO photos. Also, online media who spread fake news and frequently modifies images can ELA prove to be useful against. ELA is specifically designed to provide quick results and uses algorithms that measure changes in the compression of an image (Error Level Analysis). If there are visible differences in the image after an ELA image analysis, it has probably been digitally modified.

On the ELA website fotoforensics.com you can upload an image or refer to an image URL and put ELA on the error-level analysis immediately. The result will appear a few seconds later on the site with the original image at the top and the results image at the bottom. The resulting image reveals possible differences in the JPEG compression. With JPEG images, the entire image must be about the same level. If a section of the image is at a significantly different error level, it indicates possible digital modification.

Identical edges must have similar brightness in ELA results.

All edges with high contrast should resemble each other with the same intensity. And all edges with low contrast should resemble each other with the same brightness. With an original image, low contrast edges should be almost as bright as edges with high contrast. As for brightness and contrast, the image should be almost the same.

Error Level Analysis of original JPEG image (1)

An unedited *.jpg picture (above) downloaded directly from the camera will not give any special visible impact. It will typically be a very dark picture with weak indications of the details in the image. There will be no big differences in brightness and contrast.

Editing in Photoshop reveals digital modification

However, if you load the same *.jpg picture into Adobe Photoshop and start editing it, put a clip of a helicopter from another image taken with another camera into the unedited image, you begin to seriously create visible tracks.

Fake helicopter in cigar-shaped UFO image

The helicopter in the picture above comes from a picture I took with my Canon EOS 60D. It is inserted into the photo with the Sun Halo and the cigar-shaped UFO taken with a Sony A-100 camera. As expected, the helicopter separates clearly from the rest of the image in ELA results below. The edges of the helicopter have a high contrast and brightness in relation to the surrounding image. Here ELA’s photo analysis revealed digital image manipulation.

Fake helicopter in cigar-shaped UFO image Error Level Analysis

ELA’s photo analysis has thus discovered differences in the compression of the image. ELA can see that the inserted helicopter has a slightly different profile regarding compression. It clarifies it in the image with the slightly higher contrast and brightness (error level) around the helicopter.

What ELA photo analysis does not detect

There are also things that ELA photo analysis does not detect. If we take the original photo with the helicopter and load it into Photoshop and make a copy of the helicopter and put it in the same image it comes from, ELA will not see that helicopter No. 2 is false. It is because the compression of the original photo is exactly the same as the clip by helicopter no. 2. The clip comes from the same photo, so ELA detects no difference here, no digital image modification.

Real and "fake" helicopter together.

In the picture above, the fake helicopter is already inserted in the picture to the left. The one to the right is genuine. And the ELA image analysis below does not provide any useful results. I would be able to insert an entire armada of helicopters without ELA would discover the digital image modification. My fraud would go under the radar here.

Real and "fake" helicopter together. Error Level Analysis.

However, common sense would probably reveal the deception since it would seem suspicious if a whole fleet of helicopters had the exact same angle in the photo. But what if I take a copy of the real helicopter and rotate it 90 degrees and insert it again?

Helicopter Error Level Analysis result 1

Again, ELA is not able to see the fraud. Despite the clip with the helicopter is turned 90 degrees, it still reveals no differences in the compression. We need to have a different approach if the Error Level Analysis should be able to detect the fraud.

Digital image modification gives conflicting results

Further digital modification in Photoshop with the fake helicopter enlarged and now twice as large as the real heli, unfortunately, gives a contradictory result, see below. The real helicopter to the right in the picture now stands out more than the fake helicopter to the left in the picture. You will mistakenly believe that the false heli clip is genuine and the real is false…

Helicopter Error Level Analysis result 2

I actually expected the compression in the enlarged helicopter clip to give clear results, but in this case, the result was the opposite than expected. Only when I reduce the size of the false helicopter in the image, ELA reacts as expected, see below.

Helicopter Error Level Analysis result 3

The clip with the diminished helicopter to the left in the picture above has a higher contrast and brightness (error level). It obviously does not fit into the rest of the image. We now know the smaller heli has another origin than the original photo…

Helicopter Error Level Analysis result 4

The picture above (same as the article’s intro image at the top) showing a whole fleet of fake helicopters, visualizes the algorithm behind the error level analysis. Each time the cut of a helicopter decreases in size, the contrast, and brightness increase.

Interim conclusion

ELA image analysis cannot reveal all facets of digital image modification. A JPEG image can easily be modified in Photoshop to an extent without the photo analysis of ELA detecting it. However, we can conclude with the last two images above, that because there are differences in contrast and brightness, the photo must be digitally modified and possibly falsified.

If you feel you can not trust the Error Level Analysis (ELA), you can take a look at the photo’s metadata. It contains information that changes every time a photo is edited in and saved by an image editor like Adobe Photoshop.

The image´s digital fingerprints with EXIF metadata

If you load an unedited photo into Adobe Photoshop and just save it under a new file name without editing it, a subsequent image analysis with ELA will not reveal anything beyond the usual. The image will resemble itself as before it was saved by Photoshop – dark and no big differences in brightness and contrast.

However, a picture will lose some of its quality when it is saved again in an image editor and the compression of the image will, therefore, be slightly changed. This can be seen by the whole picture being slightly brighter in an Error Level Analysis. But the image metadata better known as EXIF data will be able to tell about a photo’s origin. It accompanies the new version of the image (as long as it is not compressed too hard) and is also read by the ELA photo analysis.

EXIF is an abbreviation for Exchangeable Image File Format. These are technical metadata (information) about an image, such as the camera model used, resolution, width, and height of the image, how you have set your optics (fx focal length) and more. And is an image editing program such as Adobe Photoshop used, it will appear from the image’s EXIF data.

Today we have another image data standard called XMP (Extensible Metadata Platform). It is a recently adopted format created by Adobe in 2012 and will also be displayed in the ELA image analysis. It actually shows the same as EXIF data does.

Images on the Internet are compressed

Most images on the internet today are compressed. The reason is obvious. The less a picture fills in the amount of data, the faster it is loaded onto a website. And since smartphones have a fairly low capacity in terms of bandwidth, compression of photos, video, and audio becomes a must to cover the needs of mobile users for fast information on the go.

Nobody wants to sit and wait for minutes before a website fully loads. Loading a website should go quickly, and any webmaster who is ahead on the beat would compress the images before they are posted on a website. The pictures are typically compressed to the utmost, without losing quality. But with optimal compression, the additional EXIF metadata will be lost.

All of my photos on this website are compressed for fast loading. Therefore, EXIF metadata can not be displayed with ELA’s analysis tool or any other EXIF viewer

Basic image metadata

The example from one of my UFO articles shows the analysis of a compressed image. Here, there are no EXIF data, just some basic information. You can view this data by clicking “Metadata” in the box at the top left of the ELA analysis results page

The basic metadata for the example above does not even reveal a date of creation. The line with CREATOR: gd-jpeg v1.0 (using IJG JPEG v62), quality = 82 refers to the component of the underlying PHP code on my WordPress website that has helped to render the image into the browser.

CREATOR: gd-jpeg v1.0 is not a program but a library or protocol if you like, see GD Graphics Library. And IJG JPEG v62 refers to the type of compression, see IJG

CNN image reveals digital image modification

Since September 11, 2001, I’ve been aware of CNN modifying both image and video before it reaches the public. For a news media like CNN, it’s important to look good at the public. Therefore, CNN’s images and videos will always review image processing.

The article here shows a picture of Boeing’s new A380 aircraft type with a beautiful blue sky as a background. Right-click on the image and choose to Copy the URL and ask ELA for an image URL analysis. The result of the analysis will reveal obvious digital image modification…

The result image shows a background that is almost black all over. The plane is “sticking out” in a graphic sense. You can simply see the A380 is a separate clip and the background is a separate clip. Here you have obviously wanted a different background.

Although the image has been subjected to compression and has no EXIF data, it obviously reveals digital image modification. So in some cases, an image analysis with ELA can quickly and easily reveal manual modifications.

Final conclusion

ELA’s photo analysis cannot reveal all facets of digital image modification. If a photo has been edited and modified with components from itself, without clips from other photos taken with other cameras, ELA may have difficulty detecting possible digital image modification. And if an image has been subjected to hard compression, the included EXIF/XMP data will not be visible or reveal anything.

So you risk being left with nothing. And if you can not perform a proper image analysis, you can not conclude anything. You should always have at least a copy of the original photo and not a compressed image from the internet. If an original photo cannot be obtained, I can not in my optics perform a proper photo analysis.

What ELA’s image analysis nor will see, is, if you hang an upturned plate resembling a flying saucer in a fishing line, take a picture of it and claim you have a photo of a UFO. ELA is only effective against digital image modification, not real things that are physically present in a photo.

If an image after analysis on the ELA website shows particular differences in brightness and contrast, the picture has undoubtedly been digitally modified.

However, with ELA, you can always stick to one rule of thumb! If an image after analysis on the ELA website shows particular differences in brightness and contrast (as with the image from CNN), the picture has undoubtedly been digitally modified.

Use the free image analysis tool wisely

After each analysis, ELA offers a direct link to the result of the analysis so you do not need to use the free analysis tool again for the exact same image the following day. A direct link to the ELA result can be found at the bottom left of the ELA results page. Remember to use this link in the future. If you abuse the tool, you risk becoming blacklisted.

The direct link in this case indicates the result of an image analysis of one of my UFO photos mentioned in the article Cigar-shaped UFO and airplane in Sun halo above my hometown Odense (Denmark) back in 2009.

ELA results images on this page

ELA guides (tutorials)

Have you already been to the photoforensics.com website where the ELA analysis tool is available, you have probably discovered how much information there is. Most people would fall backwards and abandon the business to understand all aspects of photo analysis. But somewhere you have to start and you can use the tutorials page as well as take up the challenge and solve some tasks or read their FAQs

I use ELA’s image analysis tool sometimes when some users on Twitter try to fool me with their homemade UFO images. Often it is sufficient simply to enlarge the images and reveal their home-forgery. In most cases, the search for digital image modification with error level analysis has been unnecessary. But it is very good to have if you need to investigate for digital image modification.

Dr. Neal Krawetz critical inputs

I was lucky to meet Dr. Neal Krawetz in a Twitter discussion about a Mars image shared by Twitter user @AntonioParis

Dr. Neal Krawetz and @hackerfactor maintain the foto forensics service after Pete Ringwood, which retired his site in 2012. Dr. Neal Krawetz and @hackerfactor continue this free service, which he should be thanked for.

Unfortunately, he was not satisfied with my article here.

Since I am not really a super expert in this field, but Dr. Neal Krawetz is, it´s easier for me to just show you his critical inputs here. You can then draw your own conclusion based on Dr. Neal Krawetz inputs here.

Actually, it´s my own quotes…

Thanks for that correction!

Yes, I am still learning…

Thanks for that correction!

Thanks for that punch!

So now everything should be in its right place…

Related reading

Or go back to category UFO research or homepage Real UFO pictures

Spread the love

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.