Integrating OpenAI Code Analysis for identifying harmful code

[ZanattaMichael]

Description

This pull request adds 1 new commands to PowerShell Protect for integrating OpenAI into the platform. The new commands New-PNPAICondition with 2 conditions added. These are: AIisBestPracticeCondition and AIisSuspiciousCondition.

New Commands Added

1. It takes in five parameters, including Property, AITemperature, AIRating, APIKey, and ContinueOnError. It then creates a new Condition object with these parameters and writes it to the output. The SecureStringExtensions class is used to convert the APIKey parameter to a base64-encoded string for added security. This cmdlet is part of the PowerShellProtect.Cmdlets namespace and can be used to integrate OpenAI into PowerShell Protect.

New Conditions Added

1. AIisBestPracticeCondition: This condition checks whether an AI model adheres to best practices for training and deployment. It evaluates various aspects of the model, such as its accuracy, fairness, interpretability, and security, and returns a boolean value indicating whether the model meets the best practice criteria.

2. AIisSuspiciousCondition: This condition checks whether an AI model exhibits suspicious behavior or is vulnerable to attacks. It evaluates various aspects of the model, such as its inputs, outputs, and internal workings, and returns a boolean value indicating whether the model is suspicious or not.

Usage Example

# Create a new AI condition
New-AICondition -Property isBestPractice|isSuspicious -AITemperature -1|1  -AIRating 0.5 -APIKey SecureString -ContinueOnError

Testing

Initial Testing, however AI Testing (with mocking) is not functional.

Documentation

None so far.

Contributors

• Michael Zanatta