APPLYING SOCMINT TO EXTRACT CYBER THREAT INTELLIGENCE FROM THE RUSSIA-UKRAINE CONFLICT

Abstract

The paper applied SOCMINT (Social Media Intelligence) techniques to discover cybersecurity-related information from the contemporary Russia-Ukraine conflict. Using open-source tools and APIs, datasets created were assessed through topic modeling, thematic analysis (word cloud), Logit function, and neural network classification. The topic modeling and word cloud yielded trifling insights, but Logit and neural network classifier, MLP, suggested statistically significant features that were important to the outcome of the tweets with reasonable accuracy of 91%. Through the use of synthetic data (GaussianCopula) and feature selection(stepAIC), the model was extended to improve accuracy, which resulted in 96% accuracy, though, such competent performance requires further investigation. While deciphering the right intelligence is a challenge due to the unruly nature of social media, this nascent technique can be helpful with the proper framework and approach.