APTIMA INC — Department of Defense SBIR Phase I: Efficient and accurate indexing is fundamental to the Big Data enterprise, but traditional

Description

Efficient and accurate indexing is fundamental to the Big Data enterprise, but traditional indexing techniques are often foiled by noise and missing information. Aptima proposes CERTAIN (Certainty Enrichment via Relational and Temporal Analytical Indexing of Networks), a solution that advances several areas of large dataset indexing and applies them to the domain of cyber analysis. We will develop an enhanced indexing approach that will retrieve data that traditional indexers would miss. Our enriched indexing approach will rely on inference algorithms that extract high level features in temporal and relational datasets. In the temporal domain, we will detect patterns of temporal behavior by learning categories of records that share qualitatively similar characteristics. In the relational domain, we will detect subnetwork primitives that highlight regions of a dataset that share similar properties. We will demonstrate how inference-enriched indexing improves the retrieval of relevant data by intelligently managing noisy datasets. Our research and development effort will focus on a cyber dataset, but our approach will extend to other data sources that have temporal or relational components.