Determining security vulnerabilities based on cybersecurity knowledge graphs

A system for retrieving text identifiers of documents used by a large language model to determine security vulnerabilities based on cybersecurity knowledge graphs, the system comprising one or more processors and one or more non-transitory media storing program instructions that, when executed by the one or more processors, perform operations comprising:

• receiving a query indicating a vulnerability from a user with a user security access level;
• in response to the query, determining a set of most similar documents associated with a set of unique tags by:
• accessing a cybersecurity knowledge graph tuned by a segmented corpora of cybersecurity data, wherein the cybersecurity knowledge graph delineates probabilistic relationships between at least two documents of a set of cybersecurity documents of the segmented corpora, and wherein each cybersecurity document of the set of cybersecurity documents comprises a unique tag, a security access level, and a vulnerability directive comprising data for quantifying cybersecurity vulnerabilities;
• and
• determining the set of most similar documents based on the cybersecurity knowledge graph;
• generating a vulnerability score using vulnerability directives of the set of most similar documents;
• determining whether the user security access level satisfies a set of criteria based on the security access level of each document of the set of most similar documents;
• in response to a result indicating that the security access level of the user satisfies the set of criteria, associating a set of text identifiers of the set of cybersecurity documents with the vulnerability score by tracing the unique tag of each document of the set of most similar documents determined to be most similar by the cybersecurity knowledge graph to determine the set of text identifiers for the vulnerability score;
• and
• displaying, on a user interface, the vulnerability score and the set of text identifiers.