How to verify photos and videos on social media networks
Just as propaganda wasn’t born alongside the Internet, images were being edited well before the advent of Twitter and Photoshop. To take just one example, the Soviet Union regularly erased disgraced political leaders from photos, even if the results appear woefully amateurish compared to the advanced photo-editing techniques used nowadays.
What has changed, however, is that almost anyone can now make, upload and share misleading photos or videos. Social media has effectively given anybody who has an interest in spreading fake images online the means to do so with a few clicks of the mouse. In France, the issue got widespread attention in the aftermath of the Charlie Hebdo massacre. In the hours that followed the shooting, dozens of fake photos were already flooding social media networks. The message conveyed was always the same: "They’re lying to you!" No matter who they blame for these supposed "lies"– Jews, freemasons, the United States, and sometimes all three – what these Internet users want is to sow doubt, and in doing so, discredit the work of journalists.
Is it ever possible to be 100% sure ?
It’s important to point out that it’s not always possible to say with absolute certainty that an image is fake. You could, for example, use these tips to assert that the date is wrong, or that the details shown in the image don’t quite match up with the location given in the caption. Journalists often spend hours verifying photos and videos to help editors decide whether or not to broadcast them. So, for example, even if it’s impossible to pin down the exact date a video was filmed, editors might still decide to broadcast it if they are sure that the scene depicted is authentic.
Two types of analysis
There are two complementary approaches to checking an image’s authenticity. The first involves carrying out a "technical" analysis. In concrete terms, that means extracting data stored in the video and photo files. The second involves analysing the content by beefing up the traditional fact-checking process with methods specific to social media.
There are no shortcuts, because there isn’t any software capable of checking if an image is fake. Using social media to investigate the authenticity of user-generated content is a skill that can take years to perfect.
We’ll begin with some basic tools before moving on to more advanced techniques.
Step One: When was the image taken?
Even with modern photo-editing software, it takes times and effort to create a fake image, and even more to make it look credible.
The Chinese press is packed with examples of hastily edited images. The photo below is a case in point, taken from a publication in the city of Hangzhou.
Since editing a photo is a fairly complicated process, many Internet users resort to a far simpler method. They take an older image out of its original context and link it to a recent news story. The photo below is one such example: it sparked an outcry when it began circulating online not long after a deadly stampede killed thousands in Mecca in September 2015.
Internet users of all stripes brandished the photo as proof that Saudi authorities were using bulldozers to clear away bodies. No bodies were clearly visible in the video. After investigating, our journalists were able to show that the photo was actually taken in 2004 after a similar incident had taken place. Even then, it’s unlikely that the bodies had been moved by bulldozers.
If you spot that a photo is older than its caption, your first instinct should be to put it through Google Images or Tineye. These tools will reveal any previous occasions on which the photo has already been published online.
This photo claims to show a young victim of the war ravaging eastern Ukraine.
But while this tool is certainly useful, it does have its limits. It can sometimes miss a photo's publication history. Even if a Google Images search turns up nothing, that’s by no means proof that it has never been published online. Even the American search giant isn’t fail-proof.
When it comes to videos, however, there aren’t any tools on par with Google Images for checking a video’s publication history. With the help of YouTube, Amnesty International has set up an online tool which can be used to check a video URL.
If the very same video was posted at another date on YouTube, the tool will find it. But the downsides are glaringly obvious. For one thing, it only checks content posted to YouTube. And if the video is modified slightly – even if a few seconds have been trimmed from the start or the end – the link between the two versions is lost and the tool can’t turn up any results.
Here again, technology can’t solve all your problems. EXIF data is often lost when photos are posted on websites or uploaded to social media networks. The information can also be lost when an image is modified in Photoshop. It’s therefore crucial to try and find the original image file. If it was sent directly by email, it should contain the EXIF data captured along with the photo itself.
UPDATE : Some tools for analyzing metada now also work with videos. If you upload a video on Jeffrey’s Exif viewer, the tool can sometimes tell you the date and time it was filmed. But, just like for photos, most social media platforms erase videos’ metadata. Therefore, this technique will only work if you have the original video (sent to you by email, WeTransfer, Dropbox, etc…). If the video was posted on YouTube or Facebook, the metadata will have been lost. On Twitter, from the tests we have run, it appears that only videos posted in native format will retain metadata. Some files sent via Whatsapp can also be analyzed using Jeffrey’s Exif viewer, but most of the time, Whatsapp will eraser metadata as well.
Here’s the snag: EXIF data can be altered by anyone with their heart set on misleading you. In practice, though, few Internet users have the technical know-how to go so far.
People can lie about an image’s date, and they can lie about its location, too. Many fake photos are simply taken in one country and published as if they were depicting an event taking place in another. To avoid this pitfall, many journalists carry out location-specific searches on social media networks. The aim is to pull up only those photos taken near the events unfolding by ruling out Internet users posting on Twitter a thousand miles away. There are many tools – some free, some not – designed to track the whereabouts of an Internet user posting messages on social media networks (Yomapic, Echosec, Gramfeed, SAM Desk, Geofeedia, to name but a few). Twitter’s advanced search engine is also quite a handy tool: https://twitter.com/search-advanced?lang=fr.
Advanced Twitter Search
Tweetdeck, a tool for managing personal Twitter feeds, also lets its users add location-specific codes when carrying out searches (for example: geocode:44.467186,-73.214804,200km). It’s not as complicated as it sounds, and you can find out more here.
Geolocalisation can reveal more information about an image, but it comes with a catch. Someone living in Yemen can easily post a photo to Twitter that he’s received by email from somewhere else. As a result, if you search for photos on the conflict in Yemen, the geolocalisation tool would cause the image to show up in your results even if it was taken in another country.
But don’t lose hope just yet. Despite the numerous drawbacks already outlined in this article, a technical analysis becomes formidable when it’s coupled with investigative journalism. The aim here isn’t to remind you of the basic principles needed to fact-check information, like cross-checking sources and the five Ws.
Instead, let’s concentrate on what methods are at our disposal to verify images published on social networks.
The first things to look out for are details that are inconsistent with what the photo claims to show, and to ask yourselves the right questions. Here are a few examples.
This image was mistakenly broadcast by one of France’s largest TV channels, France 2. The scene was described as having unfolded in Iran, back in December 2009. A cursory glance reveals a range of details that could allow us to verify its authenticity. Are Iranian police shields the same colour as in this image? Are Tehran’s pavements painted yellow? Is that really how young Iranians dress?
The photo was actually taken in Honduras. And there’s no better way of finding out that its caption is misleading than by showing it to an Iranian, who would likely be baffled by the fact that in December, in freezing cold weather, everyone appears to be wearing T-shirts.
Another example, far more recent, relates to the migrant crisis in Europe, a favourite theme for those who wish to mislead the European public. Our team of journalists has already debunked several fake photos and videos, including this piece of footage shared by right-wing extremists.
According to the caption posted to YouTube, the video depicts violence at the hands of migrants in Erfurt, a city in central Germany. Two crucial details should raise eyebrows though. For one thing, the assailants attacking the police car can be heard shouting in perfect German, which is surprising for a group of newly arrived Syrian or Afghan migrants. Next, a quick Google search is all it takes to see that the blue vehicles used by Erfurt’s police don’t match the green ones shown in the video. The footage was actually shot in 2011 in Dortmund. In a terrible twist of irony, the men wreaking havoc are actually neo-Nazi activists.
Time for a pop quiz! Look carefully at the photo below. How would you verify where it was taken?
Google Map and Google Earth can be useful for carrying out in-depth analyses. Keep an eye out for small details, like a bridge in the background, or a half-hidden signpost, that could confirm an image’s location. The online community ‘Bellingcat’ regularly carries out investigations along these lines, even calling on other Internet users to get involved.
Before even comparing the footage to Google Earth images, or taking a listen to ensure the sounds are consistent with the events supposedly being depicted, take a look at the user's publication history on YouTube. The first thing that stands out is that this particular channel has been hosting videos filmed in Talbiseh for several months. We already know that this town has been hit by Russian air strikes. That's reassuring, but not enough in itself to verify that the footage is authentic.
Researching Internet users is possible on social networks like Facebook, Twitter, and YouTube because they keep track of users’ activity. The growing popularity of instant messaging apps like Whatsapp and Viber is making our work more complicated. These apps provide almost no information about users who share images. They are identified only by their phone numbers – which allows you to see what country they live in, or at least bought the phone in – but it’s impossible to see what they’ve posted previously or who their “friends” are. Moreover, Whatsapp and Viber usually erase all the Exif data on images they host. This makes verification extremely difficult.
This real photo of Osama Bin Laden, alive, was blended with a photo of a dead body in Iraq. This photo demonstrates why, when in doubt, it's necessary to read the comments posted by Internet users. Someone will likely spot something that you’ve missed.
Then, listen to the words. On social media, you'll always find someone who speaks the language you're after. You probably don't speak Urdu or Lingala, but someone on social media will. What's more, most people will offer to help you if you ask them nicely - trust us on this one. Make use of social media networks to ask for translations of image captions or user comments. This method is far more reliable than Google Translate.
Relying on Internet users who we don't know or trust has its limits, however. That's why it's important to create your own network. At FRANCE 24, since 2007, we've put in place our very own network of Observers that nowadays numbers more than 6,000 people scattered across the globe. These citizen journalists work together with our team of professionals to cover news events. Thanks to this network of collaborators - made up of people who we know and trust - we can verify news stories quickly and efficiently.
Here's just one example. In October 2009, our team received a photo purporting to show a killing in broad daylight in Conakry, Guinea. Very few journalists are based in this country, so we couldn’t rely on professionals to verify this information.
Our team sent these images to several Observers based in the same city. One of them spotted a pharmacy sign and recognised the area. He headed to the scene, where by speaking to witnesses he was able to confirm that the incident had indeed taken place that very day.
Of course not everyone has their own network of Observers. But thanks to social media, everyone has the opportunity to form their own community. Whether it be through Facebook or Twitter, with time, anyone can build up a valuable network of contacts who can help you cross-check information by virtue of their location or expertise.
We've talked about technical analyses and investigative techniques, but should we also trust our instinct? Absolutely – but even that is a skill that takes time to perfect. For example, what is your first reaction after watching this video?
The footage was spread by a large number of media outlets and created a huge buzz online. But the video seemed so perfectly filmed that it looked too good to be true. If something looks too good to be true, it probably is. The video is short, the action takes place up close, and everything happens in full view of the camera. Afterwards, the man falls to the ground and flees without further ado, despite being humiliated. It's surprising to say the least. Our instinct told us it was false. From there, we were able to pick out concrete details to back up this hunch. Our team of journalists simply called the Russian bar where the footage was filmed, and in doing so uncovered the truth. It was nothing more than a publicity stunt filmed by a PR agency.
Sadly, these kind of "fake" events spearheaded by publicists are becoming more and more common. The latest big hoax: a fake "migrant" who documented his journey to promote a photo festival.
Publicity agencies don't worry about harming the credibility of media outlets or social networking sites in the process. If you slip up and spread the hoax, all the better for them.
Keep in mind that it's in the interests of many people to mislead journalists. They could be countries, political parties, conspiracy theorists, and even those with good intentions. Even human rights groups or well-meaning activists might send you a photo without knowing that it's completely fake. And if you tend to agree with their cause, that may leave you less vigilant than usual when it comes to checking the image. Always have doubts, and don't hesitate to get in touch with our team if you need a helping hand: firstname.lastname@example.org
Here are some image verification experts to follow on Twitter:
Malachy Browne, @malachybrowne from @reportedly (and of course his colleague @acarvin)
Tom Trewinnard, @Tom_El_Rumi from @checkdesk and @Meedan
Claire Wardle from @TowCenter
A coalition of experts from @firstdraftnews led by Jenni Sargent, @JenniSarge
Dhruti Shah, @dhrutishah from @BBCnews
Joey Galvin @Joey_Galvin from @storyful
Jochen Spangenberg, @jospang from @revealEU
Anne-Marie Lupu, @amloopoo from EBU