This patent covers an advanced machine learning architecture for classifying and comparing software artifacts across multiple categories simultaneously — a capability that sits at the heart of next-generation application security and threat detection.
Traditional machine learning classification assigns an input to one of a fixed set of categories. "Multiclass" classification extends this to allow simultaneous assignment across multiple non-exclusive categories. "Multimodel" approaches combine the outputs of multiple specialized models to improve accuracy and coverage. "Similarity analysis" adds the ability to find files or code artifacts that are related to known threats or vulnerabilities, even when they are not identical matches — enabling detection of novel variants and obfuscated malware.
Together, these capabilities represent a significant advance over both legacy signature-based security systems and earlier single-model machine learning approaches. The patent reflects the sophisticated AI research and engineering that made Cylance one of the most technically credible AI security companies ever built.
Stuart McClure's portfolio of patents from his Cylance years documents a sustained program of original AI/ML research applied to the problem of endpoint and network security. This patent in particular has broad applicability to application security analysis — the domain in which Stuart's current company, Qwiet AI, operates — where the ability to classify code vulnerabilities across multiple dimensions and find similar patterns across large codebases is a core technical challenge.