PRISM - Unit Summaries

Forums are hubs for malicious activity, making it essential to monitor them to stay informed about emerging threats. However, the sheer volume of activity on forums makes it impractical for humans to track and analyze every conversation. This is where leveraging a Large Language Model (LLM) becomes invaluable.

Our approach uses an LLM to generate unit summaries—concise daily summaries of conversations from individual threads on forums. For each new and relevant thread, we produce a summary that captures the key points of discussions for that day.

Unit Summary Social Network

We introduce a social network that maps the relationships between people and the concepts they discuss in forums. In this network, actors are connected to the concepts they talked about on the forums. Additionally, actors are linked to one another if they participated in the same thread on the same day. This shows not only what topics are being talked about but also who is interacting with whom, giving a clearer picture of how the community and its ideas are connected.

Each actor is enriched with attributes such as discussion spread over concepts, seniority, forum presence and their connections. Click on actors and concepts in the network to explore their details and attributes further.

This social network visualization is based on data collected up until August 2024. It is a static representation and does not update in real-time.