Overcoming Concept Drift of Machine Learning Solutions
The majority of machine learning solutions consider that the data distribution is stationary, i.e., that they do not change over time. However, when it comes to problems like malware detection, for example, this assumption does not hold up, due to the natural evolution of malign software when trying to evade detection mechanisms. The Need for Speed is something wanted in such solutions to make users and corporations less prone to vulnerabilities.
Our project aims to overcome the mentioned problem by proposing a framework capable of detecting a concept drift and taking the necessary measures to get over it. You can read more in our first paper about this project here. The framework is under development and will be made available soon.
Need for Speed is not just a contribution for the cyber-security research, but also for the entire machine learning community, for every problem involving data streams and concept drift (non-stationary problems) where it can be applied.
If you are interested in researching in this topic, join us!