Research Centre for Responsible
Media Technology and Innovation
Project number 309339
Visual Content Verification:
Observations from Faktisk Verfiserbar
Sohail Ahmed Khan
THE AIM OF THE PROJECT IS TO SOLVE THE MEDIA'S
URGENT NEED FOR EXPERTISE TO VERIFY VIDEOS
AND PHOTOS FROM THE WAR IN UKRAINE.
The secondary aim is to strengthen and further develop this
expertise for all Norwegian media, and increase the knowledge
and capability to do fact checking journalism.
Workflow
The collection and analysis of data gathered from open sources
(covert and publicly available sources) to produce actionable
intelligence.
VERIFICATION WORKFLOW
OSINT
Telegram: Russian Missiles Heading Towards Finland
• This was a video that was published on
Telegram in May.
• The video is allegedly taken near Leningrad,
Oblast.
• Claims the movement of missiles to be
reaction of Russia towards Finland’s interest in
joining NATO.
Exploring the Use of Computational Tools for Fact-Checking:
A Journalistic Case Study of Verifying the War in Ukraine
Sohail Ahmed Khan, Jan Gunnar Furuly, Henrik Vold, Rano Tahseen, Duc-Tien Dang-Nguyen
Exploring the Use of Computational Tools for Fact-Checking:
A Journalistic Case Study of Verifying the War in Ukraine
Sohail Ahmed Khan, Jan Gunnar Furuly, Henrik Vold, Rano Tahseen, Duc-Tien Dang-Nguyen
Visual User-Generated Content Verification in Journalism
mediaforensics.mediafutures.no
Sohail Ahmed Khan, Ghazaal Sheikhi, Andreas Lothe Opdahl, Fazle Rabbi, Sergej Stoppel, Christoph Trattner, Duc-Tien Dang-Nguyen
Visual User-Generated Content Verification in Journalism
mediaforensics.mediafutures.no
Sohail Ahmed Khan, Ghazaal Sheikhi, Andreas Lothe Opdahl, Fazle Rabbi, Sergej Stoppel, Christoph Trattner, Duc-Tien Dang-Nguyen
• Assists journalists/fact-checkers recognize military equipment within
images/videos.
• Recognizes 20 different tanks, and artillery vehicles (for now).
• Also recognizes destroyed/damaged tanks in most cases. In case of wrong
classification, the returned 5 most probable names typically contain the
correct answer as well, thus reducing search space for the person using the
classifier.
• Makes efficient decisions within seconds.
• Provides background information about predicted tank, gathered from
Wikipedia.
• Provides more pictures of the predicted tank allowing the user to compare
them with the query image manually.
WHY IS IT IMPORTANT?
• Assists journalists/fact-checkers in recognizing the
language being spoken in a video. In seconds.
• Tells what language is being spoken, transcribes what is
being said, and also translates it to English.
• This can be useful in verifying videos coming out from
warzones, foreign countries having languages other
than English/Norwegian.
• Reduces the manual effort in translating and
recognizing the language in a video. Currently, the
journalists have to upload the videos to YouTube and
wait for 40-45 minutes to get transcription/translation.
• Will be updated to tell what countries/regions where a
specific language is spoken widely.
WHY IS IT IMPORTANT?
OSINT involves using publicly available information from social media, websites, and news articles to gather information
about an individual or organization, or an event.
In the middle of May 2021, faktisk was tipped offabout a video claiming that Russia is sending vehicles carryingmissiles towards its border with Finland in response to Finland initiatingthe process to join NATO.
By analyzing the Russian highway network. The exact place where thevideo was recorded was found to be about an hour and a quarterdrive from Vyborg, 127 kilometers from the Finnish border.
When we geolocate – we find out where the picture is taken. We often use Google Maps, Yandex Maps or Google Earth Pro.
What do we know from watching this video?
We know it’s a six-lane highway. We can see that in the video on the right.
The post states that the video was taken near Leningrad, Oblast.
We searched this area in Google Earth Pro and startet looking for highways around Leningrad.
Looking around on Google Earth Pro, we found something that seemed promising.
We started using Google Street View and looking for elements that could match.
We finally managed to find out that this video was taken on E-18 between St. Petersburg and Vyborg. Heading north.
When we geolocate – we find out where the picture is taken. We often use Google Maps, Yandex Maps or Google Earth Pro.
What do we know from watching this video?
We know it’s a six-lane highway. We can see that in the video on the right.
The post states that the video was taken near Leningrad, Oblast.
We searched this area in Google Earth Pro and startet looking for highways around Leningrad.
Looking around on Google Earth Pro, we found something that seemed promising.
We started using Google Street View and looking for elements that could match.
We finally managed to find out that this video was taken on E-18 between St. Petersburg and Vyborg. Heading north.
So how do we know if this really is the Iskander missile vehicles?
Normally we compare the vehicles with each other and look up different elements.
But finding out whether the video was filmed after the NATOprocess had begun was another problem?
To verify this, historicalweather patterns became a useful tool in determining the timingwhen the video could have been filmed. Based on historicalsatellite imagery from the Sentinel Hub, hadthe evidence that there was snow in the area as recently as April23, 2021.
Since the video doesn’t show any traces of snow, thisinformation became critical in narrowing down the timeline duringwhich the video was filmed.
In the video, the windshield wipersof the car from which the video was filmed were running whichsuggests that it was raining that day.
checkedhistorical weather data from “weather.com” to see what days it wasraining in the area.
It was not possible to say 100 percent certain thatthe clip was from May 16. But the combination of historical weatherdata, and satellite imagery from Sentinel Hub made it likely that theclip was from May rather than from April, the same month duringwhich Finland started the process to join NATO. This verificationbecame the main story throughout on 17 May
We share all the pictues and videos in an Excel-document for all the medias in Norway. So everyone can see what we are working on.
But also to show how rough the content is.
We have verified a lot of material which shows how brutal this war can be.
… Journalists working at Faktisk.no, and around the world receive images/videos relating to the war in Ukraine typically through social media platforms and online forums. It is understood that the content shared on social media is not always factually correct and might be presented out-of-context. In order to verify if the content is indeed factually correct, journalists use a number of different techniques and computer software.
… Journalists working at Faktisk.no, and around the world receive images/videos relating to the war in Ukraine typically through social media platforms and online forums. It is understood that the content shared on social media is not always factually correct and might be presented out-of-context. In order to verify if the content is indeed factually correct, journalists use a number of different techniques and computer software.