The IAS “Stop the Bot” Challenge

04/02 By IAS Team

Looking for a technical deep dive? Skip ahead to our Methodology for a full breakdown of these results.

Last month, the IAS Threat Lab released our report on the 404bot, a domain-spoofing, malicious bot that we caught with our sophisticated fraud-detection technologies. While we’re glad to have discovered 404bot, the truth is that not all bots are bad. Some bots help put our technology to the test to ensure our bot detection and mitigation continues to be best in class.

Enter: Wombles.

Wombles

The IAS Data Science Team has been working since late 2019 to perfect Wombles, a crawling bot that allows us to objectively measure the amount of invalid traffic that IAS technology blocks — and how the technology of our closest competitor stacks up. In fact, Wombles lives entirely outside of the IAS Threat Lab as an additional testing mechanism for our own technology, with no connection to our fraud product. Previously, Wombles was used as a brand safety training bot, but our engineers realized he made a great tester for bot traffic, as well. Rest assured, Wombles isn’t just some picture-perfect solution that we engineered to prove our point. As you’ll see shortly, even our technology wasn’t able to catch it immediately, as our models hadn’t been pre-trained to detect its behavior. However, the dynamic learning nature of our tech kicked-in quickly and Wombles didn’t stand a chance.

How does Wombles work? It functions as a pressure test for verification and firewall technologies that underlie IAS and competitive products. By placing Wombles, a relatively unsophisticated bot, on the same inventory and measuring whether it is flagged as valid or invalid traffic, we can determine the difference between IAS and our competitors. (Note: all bot traffic detection should be advanced enough to catch a bot like Wombles which is tasked with doing standard things such as mechanically visiting pages, scrolling around, and using a set of residential IPs to display human behavior.) Think of Wombles as a detective – helping us fight bot traffic by constantly challenging our bot mitigation technology and trying to outsmart it. Below, we can see the results of Wombles pressure testing IAS tech as well as that same pressure test for one of our competitors:

*Please note: Figures are based on monthly projections


These side-by-side graphics show clearly that IAS has a higher cumulative block rate, by a very long way, which means more protection and less wastage of ad spend on bots who don’t shop. But blocking more doesn’t always mean blocking better: under-sophisticated and over-cautious fraud strategies can cause you to miss out on safe, high-value impressions.

In order to conduct the most accurate comparative study, we have to factor out the human traffic that some of our overzealous competitors often flag incorrectly as fraud. The best way to make this distinction is by using a combination of deterministic bot mitigation combined with dynamic machine learning tactics, a pairing that is unique to IAS. Wombles works perfectly to judge this contest because it is a machine itself, so it mimics the behavior of malicious bots and provides a control for the human traffic that often inflates the results of these studies. The data above shows that as IAS models detect and learn from bot traffic schemes like Wombles, they’re able to drastically outperform models that don’t utilize machine learning in their tactics.

[Learn more about machine learning and the  three pillars of ad fraud detection.]

__________________

The result? Across the board, IAS technology identified Wombles and began blocking the crawler within the first few hours of deployment. Competitor B, on the other hand, never blocked Wombles, clearly not using any dynamic machine learning.

__________________

Looking to learn more? We ran our study across 5 different industry verticals and found that an average of 90.8% more bot traffic was identified by IAS tech:

*Please note: Figures below are all based on monthly projections

If you would like to skip the vertical breakdown, click here to download the full methodology and request a meeting with IAS.

Auto

CPG

Finance

Pharma

Retail

Thanks to the good work of Wombles, IAS is moving the industry toward a more transparent, better-protected future.

If you’re interested in reviewing the full methodology of the study, fill out the form below.

P.S. – Our findings are not limited to bot traffic. Select the checkbox in the below form to request a meeting with IAS for a personalized brand presentation. Discover how IAS can help you reduce wastage from fraud and protect your brand from unsafe content.

Download the methodology