Tracking Fraudulent Mobile Devices with Forensic Flags
December 10, 2019
The landscape of mobile app fraud is ever-changing: dramatic shifts in fraud can happen in just days or weeks. Between Q4 2018 and Q1 2019, for example, fraud perpetuated through device farms decreased while SDK spoofing and click spamming increased. These shifts indicate that less sophisticated methods are quickly being replaced by new, more complex methods of committing fraud.
At the end of 2018, 20.4% of programmatic display ad fraud involved smartphone apps, and this is projected to increase. Whether manipulating ad display rates, creating fake ad impressions, or generating fake clicks, mobile ad fraud is big business. The financial impact of mobile ad fraud is estimated to reach $44 billion in 2022.
How We Detect Fraudulent Signals
Gravy Analytics processes over 15 billion pseudonymous mobile location signals every day. Since we first began processing location data, fraudulent signals have existed in the data supply. Bad actors are always coming up with new ways to commit fraud, so Gravy constantly monitors and quality checks incoming signals. When a new supply issue is identified, Gravy immediately investigates the problem. Later, we enhance our processing algorithms to automatically detect and filter similar, problematic signals in the future.
Earlier this fall, Gravy experienced a surge in fraudulent devices. On a single day in October 2019, 23.57 million fraudulent devices with a total of 52.1 million signals (an average of 2.2 signals per device) were found. While this isn’t a large percentage of our total supply, we detected the change relatively quickly, investigated the new pattern of fraudulent behavior, and enhanced our algorithms to flag it. Shortly thereafter, a new set of forensic flags went into production, denoting these spoofed devices.
While some of Gravy’s data customers choose to receive only verified data, others choose to receive more data with forensic flags. Our customers who are working on fraud prevention solutions can receive the extra data while others may choose to filter it out.
What Have We Learned
1. Quality is Better Than Quantity
Above all, quality is important. In the location data industry, providers often claim to have the largest number of signals or devices. Even with large amounts of data, fraudulent signals can still be present and this can cause the resulting analysis to be inaccurate. But what good is we often count the number of devices or signals without compromising on quality. Gravy finds that it can often be necessary to decrease counts to get quality data for our customers. Location data buyers need to look beyond the raw data. It is crucial to know how the data is being processed and verified.
2. Prevention of Questionable Analysis
Our customers depend on quality data. If we didn’t detect fraudulent signals and take action, then our customers are faced with a data dilemma. Ultimately, this would cause any resulting analyses or products which they create obsolete. Data-driven analyses or products are only as good as the quality of the data. Through identification of problematic signals, Gravy prevents our customers from using questionable data which would compromise their projects.
3. Detection Improves Ease of Use
By detecting fraudulent data, Gravy allows customers to be able to tailor data to their needs. This prevents customers from having to comb through the data themselves to determine what they need or don’t need. Fraudulent device data remains available for clients who are specifically working on fraud-related issues.
As we continue to monitor fraudulent mobile signals, we are constantly learning and tweaking our processes. Gravy will always be one step ahead of the fraudsters.