Quantifying Security Vulnerabilities: A Metric-Driven Security Analysis of Gaps in Current AI Standards