One of the largest problems facing social media platforms, law enforcement agencies, and general uses is the difficulty in attributing information operation accounts to specific nation state actors. When accounts are identified as being part of an information operation, it takes additional work to identify which nation-state they belong to - a process described by the Department of Homeland Security as a "painstaking process". This project was created with the goal of automating this process.
The utilized datasets were obtained from Twitter's transparency report. The current model follows the below statistics in terms of attributable actors.
When compared with predetermined labels in a separate testing dataset, the model achieved an accuracy rate of 90%.