- Dhiraj Gupta
Using Artificial Intelligence to Detect Real Humans
AI-powered ad-fraud detection and protection help advertisers build a wall of defense.
The ad fraud detection and elimination suite for the app and web platforms by mFilterIt does real-time pattern analysis of events and engagements to review fraudulent performance. In a typical anti-ad-fraud solution, a post-mortem analysis is done on the events to draw patterns and identify fraud. The engagements which show expected behaviors are marked 'Not Fraud'.
In such solutions, one doesn't require advanced data science techniques or the power of machine learning as well as any AI engine. One can analyze the data dumps and figure out genuine versus fake patterns. A logic check is required on the data, which can very well serve through any spreadsheet application, let's say advanced MS-Excel functions.
Of course, one needs to capture the right data points, which may require some integrations with the mobile or web applications to be analyzed. The game changes when one wants to go up the challenge and start building a protection layer, which can do an analysis on the go and block any potential invalid engagement. The tech requirements changed dynamically, hence the architecture.
Our solution uses data analytics, real-time computing services and cognitive logic to offer the best ad-fraud detection. The internal analysis shows that we can detect an average of 22% more ad fraud over and above the industry benchmark comparing other solutions available in the market. This is not just our analysis but jointly done with some of our anchor customers.
We also perform more than 70 checks to uniquely identify the origin of an event, including that of coming from an actual human. The set of unique data points powers all this we capture and real-time compute possible due to our lean on-prem architecture. We securely connect any application with our cloud-based solution for computation.
The layer of artificial intelligence built over it is even more critical, which equips the suite with all cognitive power. This is very core to the solution's architecture as all the thinking ability is designed and structured in it. As soon as the data starts shoring our servers, the 'brain' of the solution, based on proprietary algorithms using AI, starts drawing patterns to segregate actual versus fake, human versus BOT.
As a version 2.0 of the solution, we have taken it to the proactive stage, where we can validate the integrity of the click and base the results; it can be blocked much before the event occurs. This helps mFilterIt ad-fraud suite to protect proactively rather than post-mortem analysis only. It may sound easy to read, but it isn't elementary in architecture and implementation. As a technology architect,
I can vouch for that independently that it's one of the most complex solutions that I have come across in my career. However, we have always ensured to keep the integration as simple as possible.
Taking ad fraud to this level is another game-changer. Many advertisers who rely heavily on non-affiliate marketing would find using any ad-fraud solution less effective as payment is prepaid on such platforms and recovering the money after establishing any fake engagement is cumbersome.
With click integrity capability, advertisers can take a call much in advance where to spend. Our journey of more than 5 years at mFilterIt, where we have saved over $150 million and validated over 10 billion events for our happy and satisfied customers across the world, the brain of the solution has been one key differentiator.
Just like any human being, where you may acquire similar skills but cannot always replicate the capabilities which are powered by the thinking ability, mFilterIt's Ad Traffic Validation also thinks much ahead thanks to its cognitive power based on artificial intelligence engine, where it can distinctively differentiate between a real unique human engagement and a BOT.