AI Source Code Training Corpus Vulnerability Scanning

The patent applies preliminary action by scanning the initial corpus of source code for vulnerabilities before using it to train the AI algorithm. A test suite is executed against the initial corpus to identify potential vulnerabilities, and these vulnerabilities are then mitigated by removing or correcting the problematic code segments. This preliminary cleaning of the training data ensures that the AI algorithm learns from secure code examples, thereby generating secure code without inheriting vulnerabilities from the training corpus.

The vulnerability scanning and mitigation process is performed as a preliminary action before training the AI algorithm. By executing the test suite against the initial corpus and removing vulnerabilities in advance, the patent ensures that the subsequent training process works with clean, secure data. This preliminary action, while adding processing time, prevents vulnerabilities from being learned and generated, thereby ensuring long-term security without requiring continuous scanning during code generation.

The test suite is executed as a preliminary action against the initial corpus to identify vulnerabilities before training begins. This preliminary scanning phase uses computational resources to thoroughly analyze the training data and remove any vulnerable code segments. By concentrating the computational effort in this initial phase rather than continuously during training and code generation, the patent achieves high vulnerability detection accuracy while managing computational resource consumption efficiently.