Online Fraud Detection System Share: DIGI117 Score Awaiting client review n/a Date Published 25 September 2019 Reading Time 4-Minute Read Detect online fraud in digital campaigns, block malicious sources, and cut chargebacks with real-time fraud analytics. DIGI117’s team has developed a web-based online fraud detection system that identifies malicious activity in digital advertising campaigns and blocks suspicious traffic sources, using big data analytics, cloud platforms, and machine learning. Context Online fraud has become a real headache for mobile advertising networks over the last few years. Fraudsters use malicious tools, patterns, and malware to manipulate with mobile ads and traffic quality, causing tons of reputational and financial damages to ad networks and ultimately reducing the growth potential. Using an effective anti-fraud solution is one of the top priorities for any advertising network that wants to supply high-quality leads to its customers and secure its ad revenues in the long run. MVP Development Process The project included custom MVP development with product design and prototyping, AI machine learning for analyzing fraud patterns and anomalies, and high-performance database management for large datasets. DIGI117’s team provided the client with a detailed project estimate to specify the resources needed to create an MVP tailored to their needs, covering desired features and components for ad fraud protection. Following the client’s approval and acceptance of this estimate, our team got started with MVP development. We broke down the development process into several phases encompassing a variety of activities, namely: Making hypotheses and predictions about the must-have features Building a product backlog that reflects the client’s technical and business requirements and shapes the project vision Creating graphic design and sketches, including product branding, HTML slicing, custom themes Developing core features and functionalities, including user management, traffic tracking, fraud analytics, database management system Delivering additional features, including artificial neural network module, API integration, file export, UI theme customization Sharing interactive prototypes with user groups to demonstrate new features and optimize UX DIGI117’s team included a project manager, data analyst, four developers, graphic designer, and two QA specialists. The team followed the two-week sprint process throughout the project course to build the requested feature set, collect regular user feedback, and customize MVP according to the client’s needs. Solution DIGI117’s team has developed a web-based mobile ad fraud detection system that identifies malicious activity in ad campaigns and blocks suspicious traffic sources, using big data analytics, cloud platforms, and machine learning. The system featured a minimum viable set of features allowing our client to fight against mobile fraud and optimize ad spend quickly and effectively. The following features have been implemented in the MVP: Frontend: customizable visual and tabular reports, user authentication and management, navigation management, notification system, file export, server-side rendering SSR. Backend: microservice system architecture, custom and third-party IP blacklists, raw data aggregation, set of real-time and historical fraud detection algorithms, Google AI integration. Database and cloud DevOps: high-performance database management, multiserver environment, big data analytics. The system identifies most types of mobile ad fraud and anomalies, including: Botnets, device farms, device emulation, VPNs & proxies Click spamming, click injection App SDK spoofing and install hijacking Incentivization, duplicates, invasive ads Mobile attribution anomalies such as CTIT anomaly, wrong app versioning, abnormal device distribution It allowed using default and custom parameters for real-time fraud detection, such as SubIDs, ClickIDs, SiteID, AppID, as well as export reports by CSV or API. Technologies Early prototypes of the mobile ad fraud detection system with the core functionality became available for the client four weeks after the project kick-off. The client accessed remote Amazon-operated staging servers to try the system immediately and provide ongoing feedback. The full-featured version has been delivered and accepted in twelve weeks. The system was designed as a cloud-based microservice backend architecture deployed with AWS. It utilized default progressive page-load together with server-side rendering that improved UX. A special emphasis was made on the fast processing of big data for real-time campaign analytics and reporting. Results Digi117’s team has built an AI-powered ad fraud detection solution that detects and blocks non-human traffic in real-time. The system identifies traffic anomalies and prevents malicious activity in ad campaigns. After two months of using the system, our client has achieved these results: Refine ad inventory by blocking fraudsters and malicious traffic sources Reduce ad losses caused by low-quality traffic by 65% Create transparent and reliable ad tracking based on real-time fraud protection and machine learning Provide customers with high-quality leads and advertisements.