このページの本文へ移動

Menu

  • About 
  • News 
  • Services 
  • Investor Relations 
  • CyberAgent Way 
  • FEATUReS 

 

Press Release

"Tapple" Introduces Fraud Detection Filters Utilizing Machine Learning Technology

Automatic detection of users under 18, fake profile images, and improved removal of fraudulent users

Media

MatchingAgent Inc. (Head office: Shibuya, Tokyo; President: Takehiro Goda), a subsidiary of CyberAgent Inc. (Head office: Shibuya, Tokyo; President: Susumu Fujita; TSE: 4751) has introduced two new systems aimed at improving the service safety of "Tapple," an online dating service run by the company. These systems, developed using machine learning technology by CyberAgent's Akihabara Lab*1, consist of a "system that automatically detects users who may be under the age of 18" and a "system that automatically detects profile images that do not meet the posting criteria."

About the fraud detection filter utilizing machine learning technology

System that automatically detects users under the age of 18

To ensure services are delivered in a wholesome manner, Tapple prohibits the use of its platform by youths under the age of 18 under the 'Regulation for Online Dating Sites *2.' Users who register are only permitted to use the service after their age is verified with government-issued proof of ID such as a driver's license.

This system uses machine learning technology to filter language typically used by young persons to achieve rapid detection of users under 18 attempting to use the service with a falsified age. A distinct advantage of this system is its ability to more accurately detect such users compared with other systems that only match keywords such as "high school student" in the posts.

System that automatically detects profile images that do not meet posting criteria

In Tapple, profile images are screened*3 to prevent spoofing and other image exploitation. Until now, profile images posted by users were all visually verified by monitoring operators. The system uses a discriminant model that utilizes multiple deep learning algorithms to automatically approve images that meet a certain criteria. Meanwhile, images judged as possibly not meeting the criteria are manually reviewed by a monitoring operator who determines whether or not the content can be posted. This helps prevent inconsistent judgments from monitoring operators, making the detection fake profile images faster and more efficient.

Tapple has worked in collaboration with CyberAgent's Akihabara Lab to optimize the recommendation algorithm to improve the matching experience for users, while also utilizing image recognition and machine learning to ensure a safe and wholesome environment. The company is dedicated to providing a dating app that ensures safety and peace of mind for all users by continuing to develop services.


*1: Akihabara Lab
Akihabara Lab is the R&D organization for CyberAgent's media business. It staffs roughly 40 engineers who specialize in large-scale data processing, data analysis, and machine learning. The lab was established in 2011 to contribute to media services and company development by practically applying the data obtained from CyberAgent's media services.
 


List of public announcements from Akihabara Lab

*2: Commentary on the Regulation for Online Dating Sites (Metropolitan Police Department anti-cybercrime measures)

*3: About screening of main images

About "Tapple"

Tapple was launched in May 2014 as an online dating service intended for smartphones. As an online dating service where people use similar interests such as foods, films or sports as a jumping-off point to find dating partners, Tapple launches the relationships of some 7,200 couples each month with avid support from mainly young persons in their early twenties. Currently holding over 4 million registered members, Tapple has been responsible for setting up around 120 million matchings, making it the largest service of its kind in Japan.

URL