Fact check: How do I spot manipulated images?

Photos are used as evidence in breaking news situations or in crisis zones to illustrate a perspective of the story. In cases of "unbelievable but true" stories, there is often a factor that convinces readers that something actually occurred, and sometimes even how it happened.

Take, for instance, the burning of Paris' famed Notre Dame. It could have been a rumor spread on Twitter for the sake of going viral, but as soon as images emerged from various angles and from different sources, it became clear that it was likely true.

But in many cases, we observe that people try to sneak in old images, manipulated images, or completely staged images to make or point or to simply just get attention. The bigger a news story, the higher the chances are that a photo you come across on social media or the internet is actually a fake.

We have put together a guide on how to debunk a manipulated image.

You suspect a fake image — now what?

Thanks to editing software and apps, it has become very easy to manipulate images and craft a story you want. Therefore, one of the first things to do when you encounter an image that you suspect to be manipulated is to trace the origin of the photo.

A quick method to apply here is the reverse image search. It helps you find out if an image has been used before and, if so, when and how. So, let's say there has been an earthquake in Pakistan and people are sharing images online. You can either save such an image to your computer desktop and then upload it to the tool, or use the image URL and paste it in a reverse image search engine and find out if the image is actually from the claimed area.

You will also be able to tell if it is from the day it is claimed to have been taken or is actually days, months, or even years older. By looking up the first upload you can also tell, whether an image looks the same the first time it was uploaded and the way you see it now. This way you could notice manipulation.

Which tools can I use for a reverse image search?

There are various free reverse image search tools that can be used. Here are three that have very good functions.

Google reverse image search: The most well-known tool is Google's reverse image search. There are two major advantages with this tool: Google has collected the most data over the years due to the search engine's popularity and reach. So, you have a really big database that you are searching in. Another big advantage of Google is that it has an integrated facial recognition tool that makes searching for images of people much easier. It also allows you to sort the images by size. That comes in handy when you want to zoom into an image to look at details and the version you have is too bad and becomes blurry.

Yandex and TinEye: Two good alternatives to use — and you should never rely on just one tool to do the job thoroughly — are Yandex and TinEye. Yandex is a Russian alternative to Google but can give you different and sometimes better results when Google won't, especially when it comes to Russian images.

TinEye's major advantage is that it lets you sort the results by date. So, with this tool, you can find when an image was uploaded. TinEye, like the other tools, cannot tell you when it was shared on Facebook, Instagram, or messaging apps for the first time, but rather when it appeared on a website or Twitter, for example.

There is also one tool that combines all relevant image search engines and can be installed as a plugin to your browser: InViD/WeVerify plugin. The plugin has some additional features that are great when it comes to verifying images as well: an image magnifier, metadata analysis (if available), and a set of forensic tools.  

Old images or manipulated content?

We used reverse image search to find out if images being shared by both sides on the Israel-Gaza conflict in May of 2021 were manipulated or old. And we found examples for both.

 

One case of images from different and older conflicts was this one showing a boy standing in between debris and rubble of destroyed buildings. In this case, the image does show a Palestinian child and the surrounding has not been manipulated, however, the image is from October 19, 2014, according to Getty Images.

According to the news agency, the Palestinian boy is walking through the Shejaiya neighborhood of Gaza City, which was destroyed during the 50-day conflict between Israel and Hamas in 2014. 

Since then, the picture has been wrongly used by some media outlets and on social media to illustrate the airstrikes in Syria. 

There are cases where real images are manipulated with software like Photoshop to tell a different narrative. Here, image forensics can be used to help detect if an image has been manipulated or tampered with. Forensically or Foto Forensics is one such tool. It analyzes an uploaded image for clues that it has been edited or altered. Fun fact: One of the most popular items to be manipulated into images is actually a shark. 

The most common feature of this tool that is used is Noise Analysis. In an altered image you can see a difference in pixel density. Whenever you add something to an image you will most likely never be able to achieve the same pixel density as in the original. The noise level increases as an image gets altered again and again. Another helpful feature is the Error Level Analysis (ELA). ELA highlights differences in the JPEG compression rate. 

We used this tool to find out whether the image of the supposed "time-traveling" Greta Thunberg that went viral in November 2019 had been manipulated to include her — or whether it was actually a real, archival image of a girl who just happened to look exactly like the environmental activist.

Be careful with screenshots

But you should not only be cautious with images. Screenshots can also be used out of context or manipulated. At the beginning of December, for example, a screenshot of an alleged report by Deutsche Welle circulated, which showed thousands of anti-vaccination campaigners in Germany kissing publicly to demonstrate against coronavirus restrictions. However, the shared photo was old and from Chile, while the report was never published by DW — so the screenshot was a fake.

Screenshots of posts on social media platforms are not convincing evidence that those statements were actually published by that profile. The text of a tweet, for example, can be easily manipulated via source code without having to use an image editing program. Such manipulation does not get published in the original profile, but rather distributed as screenshots. You could add any possible statement to any account. And it looks deceptively real.

Even the best tools have flaws

Finally, keep in mind that no tool is perfect. It is therefore highly recommended to always do multiple searches with different tools, and then compare the results. 

If you have doubts about a particular image, apply the image verification tools and read up on the story behind it. If it seems too good to be true, it probably is. So when in doubt, be skeptical — and avoid sharing the image until it can be verified. Otherwise, you've become another person who's fallen for manipulated content.

More on how to spot misinformation:

• How do I spot fake news?
• How do I spot fake social media accounts, bots and trolls?
• How do I spot state-sponsored propaganda?
• How do I spot a deep fake?

 

Edited by: Stephanie Burnett