Wi-Fi Map In-App Fraud Detection
Enhancing Wi-Fi Map's Credibility with Machine Learning-Based Fraud Detection
problem
We were approached by Wi-Fi Map, the world's largest crowd-sourced Wi-Fi community with over 50 million app downloads from Google Play market. Wi-Fi Map app shows you the hotspots nearby and gives you the passcode to access them. Their app, designed to simplify Wi-Fi spot usage by providing accurate network information and login access details, had encountered a significant challenge. It had come to their attention that certain users were submitting fraudulent data entries within the app. This issue threatened the overall quality of the product and undermined the user experience. Despite providing valuable Wi-Fi spot information to millions of users worldwide, fraudulent entries undermined the application's credibility and reliability.
problem
We were approached by Wi-Fi Map, the world's largest crowd-sourced Wi-Fi community with over 50 million app downloads from Google Play market. Wi-Fi Map app shows you the hotspots nearby and gives you the passcode to access them. Their app, designed to simplify Wi-Fi spot usage by providing accurate network information and login access details, had encountered a significant challenge. It had come to their attention that certain users were submitting fraudulent data entries within the app. This issue threatened the overall quality of the product and undermined the user experience. Despite providing valuable Wi-Fi spot information to millions of users worldwide, fraudulent entries undermined the application's credibility and reliability.
solution
To address the issue of fake data, our task was to enhance the Wi-Fi Map app with automatic fraud detection capabilities. Leveraging machine learning techniques, our team was tasked to develop a component to identify potential fraud during the Wi-Fi spot registration process. This component would analyze user data and actions within the system detecting patterns indicative of fraudulent behavior and triggering alerts for further investigation. The manager would then receive a flag notification for a suspected fake data alert requiring further action.
solution
To address the issue of fake data, our task was to enhance the Wi-Fi Map app with automatic fraud detection capabilities. Leveraging machine learning techniques, our team was tasked to develop a component to identify potential fraud during the Wi-Fi spot registration process. This component would analyze user data and actions within the system detecting patterns indicative of fraudulent behavior and triggering alerts for further investigation. The manager would then receive a flag notification for a suspected fake data alert requiring further action.
results
Initially attempting clustering methods to distinguish fraudulent hotspot tips proved ineffective. However, after in-depth data analysis, we identified a parameter indicating user bans, which served as a key feature for training machine learning models. Splitting users into two groups based on this parameter, we successfully trained an xgBoost classifier to detect fraudulent entries. Within a month, we resolved a long-standing problem for the client, providing a reliable solution for identifying and mitigating fake Wi-Fi spot registrations. The integration of automatic fraud detection has strengthened Wi-Fi Map's credibility and improved user trust in the platform.
results
Initially attempting clustering methods to distinguish fraudulent hotspot tips proved ineffective. However, after in-depth data analysis, we identified a parameter indicating user bans, which served as a key feature for training machine learning models. Splitting users into two groups based on this parameter, we successfully trained an xgBoost classifier to detect fraudulent entries. Within a month, we resolved a long-standing problem for the client, providing a reliable solution for identifying and mitigating fake Wi-Fi spot registrations. The integration of automatic fraud detection has strengthened Wi-Fi Map's credibility and improved user trust in the platform.
tech stack
tech stack
