Artificially intelligent systems, methods and media for canonicalizing computer system logs into natural language processed representations for the purpose of data analysis

A method for canonicalizing computer system logs into natural language processed representations for data analysis to solve a technical problem of requiring specialized parsers for heterogeneous log formats from different cybersecurity tools, the method comprising:

• receiving a log file wherein the log file is received from tools having different proprietary log formats that would otherwise require specialized parsers for interpretation from cybersecurity tools including firewalls, endpoint detection and response systems, cloud security platforms, security information and event management systems, and intrusion detection frameworks;
• transmitting the log file to a cyber security purpose-based large language model;
• applying natural language processing by the cyber security purpose-based large language model to the log file to overcome a heterogeneous nature of the computer system logs emanating from disparate sources and tools, each with unique format and semantics;
• generating a plain English translation of the log file by the cyber security purpose-based large language model that obviates a necessity for specialized tool-specific integrations for data conversion and processing, wherein the cyber security purpose-based large language model generates a standardized format response wherein the standardized format response eliminates a need for tool-specific parsers by providing a unified representation regardless of originating vendor proprietary formats irrespective of diversity in tool vendors or inherent content of the computer system logs;
• canonicalizing the plain English translation of the log file by the cyber security purpose-based large language model to standardize disparate log data into a unified format and enhance interoperability of the log data across multiple system environments, including removing log-specific variable content such as IP addresses, geographical locations, ports, and protocols to distill the log data into a quintessential log signature;
• generating a multi-dimensional vector from the plain English translation of the log file by the cyber security purpose-based large language model using a paragraph embedding model wherein the paragraph embedding model generates a fixed-length vector representation that abstracts away tool-specific formatting differences to encapsulate the log data in an optimally efficient numerical representation;
• applying a cosine similarity calculation to the multi-dimensional vector;
• generating a multi-dimensional natural language alert signature from the multi-dimensional vector to enable identification of log entries that bear similarities irrespective of an originating tool vendor;
• storing the multi-dimensional natural language alert signature in a vector index database;
• applying a machine learning algorithm to the multi-dimensional natural language alert signature in the vector index database wherein the machine learning algorithm comprises Generative Artificial Intelligence models for intelligently predicting security events for preemptive cyber defense mechanisms;
• associating the multi-dimensional natural language alert signature with the log file;
• finger-printing the multi-dimensional natural language alert signature with the log file to generate a finger print;
• associating the multi-dimensional natural language alert signature with the plain English translation;
• matching the finger print to another log file to substantially condense log information thereby achieving a reduction in storage requirements and facilitating an extension in log retention durations;
• and
• including a prompt with the transmitting of the log file to the cyber security purpose-based large language model.