A Survey of Source Code Representations for Machine Learning-Based Cybersecurity Tasks

Our paper, A Survey of Source Code Representations for Machine Learning-Based Cybersecurity Tasks, provides a comprehensive overview on different source code representations that have been used in machine learning pipelines, for tasks relating to cybersecurity. Understanding the representations used allows researchers to choose or create a representation that best suits their needs.

The representation of data is a key part of the machine learning pipeline and allows models to better identify aspects of an object (in this instance, code), such that they can more effectively learn features that allows it to complete its task correctly.

We identify the representations used, cybersecurity tasks covered (as well as how they relate to representations used), programming languages that are analyzed by the tools, and lastly the machine learning models used. This work provides information about the state of the field and provides as a fundamental guide to building machine learning-based tools for cybersecurity tasks (and beyond that- coding tasks in general).

Feel free to read the paper here: https://dl.acm.org/doi/abs/10.1145/3721977