Cyber Security Solution for Online Advertisement Frauds
Abstract
Online Advertisement is a form of advertisement that uses internet pages to deliver the messages to the consumers for content promotion. Now we have almost 20 billion users over the internet, making Internet the biggest platform for advertisement. And today most of the sellers are using digital media for advertisement. As the need of digital platform increases, the types of attacks are being advanced day by day that causes many problems in digital marketing fields in terms of reputation as well as revenue.
The most prominent frauds that need to be taken care in digital advertisement are Ad frauds and more particularly click frauds. Click Fraud is a method of generating illegitimate clicks on digital ads by making repeating clicks, with the intention of generating revenue for the host website and draining the revenue from the advertiser. From 2014 to 2017 itself, the impact of click fraud has led to high revenue losses in the digital marketing sector. More needs to be done to detect these attacks efficiently and make the online advertisement realm free from these kinds of frauds.
In this project, our goal is to mine the logs of click patterns and develop machine learning models for detecting click frauds. The available logs from the customer are used to generate a labelled dataset (using the existing rule-set), which is the starting point for training the models. This project will leverage past research in anomaly detection as well as in supervised and unsupervised learning. A user’s click patterns are mined, keeping in view the current clicks as well as the history of the user. The users which falls under the suspicious category (based on a set of preliminary parameters), their click patterns are subjected to additional analysis.