Thread by @bellingcat

Creating Impact: A Year On
#StopChildAbuse – #TraceAnObject
The last article in our 2019-2020 series bringing 3 bonus case studies.
We also wrap up and quantify a year of contributions and review reasons on why should we all support @Europol's campaign.
bellingcat.com/news/uk-and-eu…

Let’s start first explaining how we tracked down three images published by @Europol and geolocated them to Mexico, Panama and Russia:

CASE 1: On flyer #G1912019 a motor bike was noticed. We analysed broken texts in Spanish, a telephone number and payment details. The incomplete puzzle indicated a potential home delivery service advert. The word “KING” led us to the food franchise “BURGER KING”. But where?

Mexico had the largest number of BK restaurants in Latin America.
A YouTube video showed a BK ad for delivery services in Mexico City. Checking Google street views for the Mazaryk BK branch, a banner matching the telephone number 1-454-KING (5464) from flyer #G1912019 was found.

A flyer sharing the same layout as seen on flyer #G1922019 was found on social media and was related to a BK branch in Michoacan. A different telephone number was noted.

1-454-KING (5464) was used by 46 BKs around Mexico City since 2013. Most BK restaurants in the area started to display new numbers as the stores were upgraded around 2015. Thus, it is very likely #G1912019 was produced between 2013 & 2015. #StopChildAbuse #TraceAnObject @Europol

CASE 2: The sports arena #G2312019 was listed and classified by Europol as probably being in Latin America. Blue seats with no numbering were observed. A practice field, a general admission section, or a stadium for an amateur league that did not assign seat selection for fans?

We found an arena with almost identical blue seats: PNC Park in Pittsburgh, home to the Pirates. When comparing the roof structures and seat distribution, we concluded #G2312019 was definitely a baseball arena, but likely much smaller than PNC Park.

We considered five baseball stadiums for each Latin American country with a recognised baseball tradition. In Panama, just adjacent to the massive "Rod Carew" Stadium, there is a smaller baseball stadium called "Leon Felipe Mota”

It was clear the chairs, structure, the light tower and background terrain of this small facility were a perfect match to those shown in #G2312019 and sought by @Europol

The small arena was inaugurated in the year 2000. According to local news from 2012, the property was managed by The Lions Club Panama who have continued organising baseball tournaments for junior leagues in this facility.

20 years of images revealed a redecoration of the stadium took place between Jan - Jul 2015. The steel structure changed from red to yellow & the steps were changed to blue. Thus, #G2312019 was most likely produced in or before July 2015. @Europol #StopChildAbuse #TraceAnObject

CASE 3: #G422019 was described by Europol as likely originating from a Russian-speaking country.

We checked old/abandoned industrial buildings in the former Soviet Union to understand the type of architecture seen in #G422019. This process helped us to define the likely type of facility we were looking for: a district boiler house (котельная, or kotelnaya, in Russian).

A small tree in the middle of #G422019 projected a shadow on the chimney stack. With this information we estimated a solar elevation angle of approximately 57 degrees.

Five cities were chosen as representative latitudes with a majority or sizable Russian-speaking population. The solar altitude in Saint Petersburg never reaches the 57 degrees mark (not even in summer), so this city was not shortlisted. The next on the list was Moscow.

Using the term “котельная москва” (Boiler House Moscow) in Yandex led us to a perfect match for #G422019 Here: bit.ly/2zd18LU . The image was taken from an off-road path adjacent to the facility. #StopChildAbuse #TraceAnObject @Europol

That’s a bit about the three geolocation case studies. Now let’s review a whole year of effort behind these investigations:

Europol posted a total of 60 new images between February 2019 and March 2020.
A total of 50 have been removed from their platform and were likely identified.
We worked on approximately 25 cases and identified 12 places and objects.

What’s the effort behind #StopChildAbuse #TraceAnObject?
@bellingcat members have devoted more than 2,500 volunteer hours to the campaign.

Looking for clues, we checked more than 2 million images on the web and surveyed the equivalent to 70,000 square kilometres on Google Earth. Approximately 1,000 documents were researched as well in the campaign

Digital enhancement and reconstruction of images have been essential in assisting visual analysis. Site sketching, detailed 3D modelling, and building shadow simulation have provided key information to geolocate and chronolocate several cases.

Links with different communities via talks, workshops, or active participation in other channels have also contributed to the investigations with the help of a growing network of supporters

With our article “Europol’s Child Abuse Image Geolocated In Ukraine: A Forgotten Story Hidden Behind A Landscape”, we have been nominated to the @EuropPressPrize 2020 for Innovation. bellingcat.com/news/uk-and-eu…

The main objective at the core of our enterprise has been to increase the chances for any one child to be rescued @bellingcat @Europol #StopChildAbuse #TraceAnObject

The fight against online abuse is much more complex than just identifying a batch of old censored Child Abuse Material. Then, why should we all participate in #StopChildAbuse #TraceAnObject from @Europol? Let's put things into perspective

In 2019, 60 million child abuse images were reported by Facebook in the US ONLY .
99% of all online CSAM goes undetected.
The global scale of the problem has been classified as an epidemic of unprecedented proportions which is overwhelming law enforcement resources

At @bellingcat, we base our work on the impact it can have on the campaign and problem:
1. Increasing the number of participants
2. Faster and more efficient identification of information, leading to the rescue of victims
3. Contribution to eliminate and prevent online abuse.

1. A Growing Network:
#StopChildAbuse #TraceAnObject relies largely on spotting pieces of local knowledge in every corner of the world.
The larger the number of people aware and engaged with the campaign, the larger the chances for objects to be identified.

@bellingcat's investigations on #StopChildAbuse #TraceAnObject have been shared to create awareness and to contribute with methods applicable to identify future cases.

By sharing notes and brainstorming on different channels, @bellingcat have contributed to emerging communities that are growing extraordinarily. An example of this is the subreddit r/TraceAnObject. bit.ly/2XTL1NH

In our workshops, we discuss online open source investigation techniques applied to our #StopChildAbuse investigations. Many of the attendees now follow and support @Europol's campaign

We have been also working with other networks to spread the word. For example, we will speak at the 2020 @NCPTF8 's Conference to help with providing access to investigative expertise in a joint mission to protect children. ncptf.org/conference

2. Rescuing Victims:
10 victims have been rescued via @Europol's #StopChildAbuse #TraceAnObject campaign. These children may never have been rescued without the help of members of the public. The rescue of even one child makes the massive collective effort more than worthwhile.

3. Better Work Flow Efficiency:
Some images might not directly lead to a victim’s identification. However, identifying an item will eliminate investigative trails and investigators will be able to focus on other leads with a higher potential for success.

@Europol is connected to @Interpol’s International Child Sexual Exploitation database.
Hence, every image identified could be linked to another series of images. While the total number of material identified could be higher, victim identification could be faster too.

4. Eliminating online Child Sexual Abuse Material:
Helping @Europol identify abuse images means that they will indirectly help eliminate abuse material via a process called hash value (digital fingerprint) comparison. We review this process in more detail in our report.

5. Preventing online abuse:
New technologies are emerging to detect online abuse. AI image classifiers scan images for nudity, age, and body motion to determine whether a content being uploaded is CSAM. The algorithms behind this technology are trained by known CSAM databases.

Deep learning techniques, such as Scene Information, help to detect objects and landmarks to locate the place of origin of an image — but these still need more development.

Without discussing the barriers in resources and technology; the database of identified CSAM we are helping to grow in #StopChildAbuse #TraceAnObject, could potentially be used to train or improve AI image classifiers and object detection to prevent dissemination of CSAM.

Recommendations for @Europol’s #StopChildAbuse #TraceAnObject are proposed in 4 key areas:
1. Better visual presentation of images
2. Better description of items base on scene information
3. Improving the way feedback is given
4. Increasing the engagement of everyday citizens