Open-Source Intelligence (OSINT) Content Processing platform
x-RD designed and developed an OSINT content processing system with a range of AI services.
Challenge
Collecting and processing Open-Source Intelligence (OSINT) is challenging due to the volume, diversity and unstructured nature of the data.
Recent events have demonstrated the value of OSINT for military and national security use cases. During the COVID-19 pandemic, analysts could use OSINT to help meet intelligence requirements working remotely. During the Russian invasion of Ukraine, OSINT provided troop movements and real-time intelligence that shaped global understanding of the conflict, and the US publicly released OSINT reports to discredit Russian false-flag attacks to justify the invasion.
x-RD was engaged by an Australian Government client to provide a range of OSINT services including the development of an OSINT content processing system for intelligence enrichment and analysis.
Solution
The OSINT content processing system integrated a range of different data types and sources and supported batch and real-time processing pipelines.
Data types including text, audio, and imagery/video were supported, and different content types were organised into different processing pipelines to extract information and metadata for population into an appropriate schema.
Imagery was pre-processed using several machine learning (ML) approaches to support super resolution, noise filtering, face and object detection, attributes estimation (age, gender, emotions) and image transformations. Text was extracted from imagery and video using ML-augmented image-to-text approaches (including OCR).
Audio was processed with speech detection, language identification, speaker ID and speech-to-text (English) using multilingual Natural Language Processing (NLP) models. The NLP pipelines included ML approaches to detect names, dates and times, phone numbers, emails, personal identifiers and mapped relations, summarised content, keywords, sentiment, and topics.
The content processing systems also conducted entity normalisation, resolution, and linking techniques to provide a high-level view or connections across the data.
Impact
The OSINT content processing systems demonstrated the automated and efficiency by why data could be rapidly processed and enriched for intelligence situational awareness and threat assessments.