EightEye

Logo
Landing Page and User Prompt
Timeline Example
Event Example
Insights Example

Video Demo: https://www.loom.com/share/257d280d9da343dabf69603aae3f0966?sid=e1cc388e-f448-49f5-b714-080bd0402850

Inspiration

Our inspiration for EightEye emerges from the complex landscape of national security, where professionals often struggle to quickly and accurately synthesize vast amounts of unstructured social media and geospatial data into actionable intelligence. This challenge hinders timely decision-making and effective response strategies during critical security events, such as public disturbances or tactical operations. Our team---with experience and interest in defense technology, OSINT, and cybersecurity---is committed to resolving this issue; thus, EightEye was born.

What it does

EightEye is a dynamic insight timeline tool that enables real-time and retrospective analysis of security-related events. The user can provide a query or prompt about a certain event or occasion that they seek to gather information or intelligence on; our tool then intelligently parses through social media datasets (e.g. Telegram) and geospatial data (e.g. SkyFi) with the use of semantic searching and NLP to construct a chronological timeline that details various important aspects of the incident, either in hindsight or in real time. By presenting a concise summary of the time, location, and key descriptors and images of the event of interest in a streamlined visual format, users can easily put together key pieces of intelligence and accurately determine next-steps. In addition to the events timeline, there are two other features that further aid the user in intelligence gathering and decision-making. Our tool has an AI chat interface that allows the user to ask questions, clarify understanding, and identify connections about the event of interest, as well as similar interests that the tool may deem relevant and useful to mention. The tool also has an "insights" panel that lists action items, suggestions, and points of consideration to guide the user through incident response and/or remediation measures.

How we built it

Our tool is obviously data-driven; for data collection, we initially rely on the proprietary Telegram and geospatial datasets provided by hackathon sponsors, as well as use of Twitter's API to gather real-time information. This data is subject to further processing (pandas, nltk, etc. for data manipulation and NLP), and we used OpenAI and Anthropic's API for natural language and image processing tasks (e.g. pattern recognition, anomaly detection, and predictive analysis based on historical data). The backend was developed with MongoDB (NoSQL database, useful for storing unstructured data), FastAPI (to host an instance of MongoDB), LangChain for custom processing, and Hugging Face for embeddings. The frontend was developed with next.js.

Challenges we ran into

Collecting and retrieving data proved to be relatively trivial, but clustering the data was a hassle. Specifically, the geospatial data was a bit difficult to work with, but we eventually figured it out. We also faced some issues with auto-generating queries using NLP, but we resolved those issues after tweaking some things and looking deeper into the APIs for the LLMs we used. The most difficult task was curating and returning parsed, labeled, and useful data to the user in a visually appealing format, which took us the most time to complete.

Accomplishments that we're proud of

We're proud of the work we've put in to address such a pressing issue in the national security and OSINT space, and we feel that EightEye is a great step in the right direction. Each member of the team was working with new technologies and frameworks at some point in the hackathon. We're also proud of our time-conscious decision-making abilities, as we made a major pivot early in the hackathon and committed to the change.

What we learned

We each learned so much about the technologies we worked with, from the intricacies of the NLP tools that we employed for semantic search to the clustering techniques we used to organize the unstructured data. Overall, the national security hackathon was a very enjoyable and rewarding experience, and we're glad to be a part of this driven community.

What's next for EightEye

We plan on using more satellite imaging data to further improve location data and inference; in addition, we hope to provide support for other social media APIs, and also identify websites and news sources as new sources of data. This will continue to the efficiency and effectiveness of our tool, providing users with an unparalleled OSINT experience.