Siri for Network Configuration Link to heading

Problem Statement Link to heading

Efficiently validating and correcting network configurations is a critical aspect of ensuring the reliability and security of cloud systems. This research explores the use of GPT-4 Turbo, a state-of-the-art Large Language Model (LLM), in identifying and rectifying network configuration errors. The study evaluates its performance on 100 flawed network configuration files, revealing a success rate of 70% without contextual information, improving to 85% accuracy through in-context learning. Future research avenues, including observed hallucinations in LLM-aided validation, are also outlined.

Contributions Link to heading

Injected faults into network configurations based on common root causes and validated outcomes with Batfish
Utilized GPT-4 Turbo to methodically detect and resolve network configuration errors, attaining a 70% accuracy rate without contextual information, and an 85% accuracy rate through in-context learning