Methods and systems for creating multi-dimensional baselines from network conversations using sequence prediction models

Abstract

Systems and methods are provided for utilizing natural language process (NLP), namely sequence prediction approaches, in the realm of network security. Techniques include analyzing network transaction records to form network sentences representative of network activity. The network sentences are formulated by regularizing transactions records using words, allowing the network sentences to represent the network activity using natural language terminology. In some cases, multiple variations of the network sentences having different sequences of words are generated to form a corpus of network sentences related to a semantics of network activity. Accordingly, an NLP-based network prediction model can be created and trained using the corpus of network sentences. The network prediction model can be trained over to identify dimensions corresponding to particular sequences of words in the network sentences, and predict an expected dimension. Using the network prediction model predictions of expected network are provided, and anomalies efficiently detected.