The system, which the researchers implemented in two days, was made available to the Information Technologies and Communication Authority (BTK) and other relevant institutions.
The software, called “Fraud Detection with Twitter Data Analysis” and currently only working on Twitter, scans tweets containing keywords such as “AFAD”, “Kızılay”, “AKUT”, “AHBAP” to ensure that the IBANs in them belong to the relevant institutions. checks if it’s not.
Account numbers that do not have permission to collect aid are marked as “suspicious” by the software and shown to the user.
“We made it accessible via a web service”
İlknur Dönmez, who works as a specialist researcher at BİLGEM, said that it was revealed that some malicious people requested money to be deposited into their account numbers by using the names of aid organizations on social media.
Stating that TÜBİTAK wanted them to make a program on the subject, Dönmez noted that they started to work quickly to identify the Twitter posts containing IBANs that do not belong to official institutions by using keywords.
Explaining that they completed and delivered the program in two days with a team of 11 people, Dönmez said, “We opened it to external access via a web service. BTK accesses the program and checks whether the tweets we marked as suspicious are really suspicious and takes the necessary action.” he said.
“We aim to reach wider social media areas in the future”
Noting that the program evaluates both the text fields and the attached pictures together in the tweets, Dönmez continued as follows:
“First, it passes the pictures through optical scanning systems and converts them into text. Thus, it finds keywords such as AFAD, AKUT, Kızılay, AHBAP in both the text and the picture. In addition, it extracts the IBANs, if any. Then it is checked whether those IBANs belong to the relevant institutions. “If there is a conflict, they are marked as suspicious by the system. Thus, the program allows the user to see suspicious tweets. 1 million 500 thousand tweets containing keywords were scanned. 1000 of them were marked as suspicious.”
Dönmez stated that the program only works on Twitter for now, but that they aim to reach wider social media areas in the future and added that they plan to improve the program with the notifications from BTK.