How to avoid fraud
How it works
Custom in-game parameters passed in the token enable the Anti-fraud system (AFS) to aggregate data about users with respect to each transaction. Information such as registration date, last password change date, total sum of payments, or game session time is used to differentiate between fraudulent users and legitimate ones as precisely as possible. The more data about user devices and behavior patterns you provide to Xsolla, the better AFS can customize its filters to detect fraudulent activity in each identified user group and across your projects.
To add custom parameters, contact your Customer Success Manager or email at csm@xsolla.com.
Examples:
- A registration date, total playtime, and user level in a game can be used in chargeback disputes to establish a user’s typical activity and detect potential fraudulent behavior.
- Dates of last login and password change are helpful when detecting hacked accounts.
Recommended parameters
We recommend to pass the following additional parameters in the token’s custom_parameters object:
| Parameter | Type | Description |
|---|---|---|
registration_date | string | Account creation date per ISO 8601. |
last_change_password_date | string | Last password change date per ISO 8601. |
total_sum | float | Total amount of payments. |
unlocked_achievements | integer | Number of achievements unlocked. |
user_level | integer | Player’s level, reputation, or rank. |
pvp_activity | boolean | Whether the player takes part in PvP (Player(s) versus player(s)) battles. |
total_bans | integer | Number of times the player was banned in the chat/forum. |
total_friends | integer | Number of friends. |
session_time | string | Average session time per ISO 8601. |
Use case
The example described in this section presents a successful case of AFS usage to significantly reduce the number of fraud cases.
A new game was launched at the end of May 2022. Immediately following the start of sales, AFS identified a significant number of fraudsters in the project: the fraud rate in the project reached 4.5% by the end of the first month after the launch.
Fraud rate dynamics by months:
To reduce the fraud rate to the acceptable range of 0.1-0.3%, we implemented the stringent AFS application for this project. As a result, the acceptance rate decreased from 96% to 81.7% on a monthly basis and to 70% on some days. This result indicates that more users underwent additional checks and difficulties making payments, particularly if their payment parameters and behavior patterns resembled fraudsters actions.
Acceptance rate dynamics by months:
To further improve the fraud defense, it was agreed to pass the following in-game parameters about the users to AFS:
- total playtime (
total_hours) - user level in the game (
user_level) - number of user’s characters in the game (
total_characters)
Using provided data, we modified the AFS filters with the new parameters between July 20 and July 22, as shown in the graph below. This led to a 15% increase in the daily acceptance rate, from ~71.4% to ~86%. Additionally, the number of fraudulent transactions reduced from 4.5% to the acceptable level of ~0.14%, and remained at this level in the following months.
Acceptance rate dynamics by days:
Found a typo or other text error? Select the text and press Ctrl+Enter.
