ML Model Validation Platform for RPA Security
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current robotic process automation (RPA) systems lack a sufficient operationalization vehicle for machine learning (ML) models, particularly in preventing the deployment of malicious code within these models.
Innovation Solution
A centralized platform for validating ML models before deployment, which performs primary and secondary validations to ensure compliance with requirements and detect malicious code, using tools like ReversingLabs or ClamAV to identify and block threats.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If third party ML models are integrated into RPA workflow, then functionality and adaptability are improved, but security risk increases due to potential malicious code
Solution Approach 1:
The patent implements preliminary validation actions before model deployment. The system performs primary validation to check model compliance with RPA requirements and secondary validation using security tools like ClamAV and ReversingLabs to detect malicious code. This preliminary screening prevents harmful models from entering the RPA workflow, resolving the contradiction by enabling third-party model integration while filtering out security risks in advance.
Solution Approach 2:
The patent introduces a validation platform as an intermediary between third-party model sources and the RPA workflow. This intermediary layer includes validation servers that execute security scans and compliance checks. The intermediary filters and screens models before they are deployed, allowing the system to maintain both functionality from third-party models and security through centralized validation.
2Reliability
If validation process is implemented, then security is improved, but deployment time increases
Solution Approach 1:
The patent segments the validation process into distinct phases: primary validation for compliance checking and secondary validation for security scanning. This segmentation allows each validation layer to operate independently and efficiently, with primary validation performing faster compliance checks and secondary validation performing deeper security analysis only when needed. The segmented approach maintains high security standards while optimizing deployment time.
Solution Approach 2:
The validation platform operates autonomously to perform security scanning and compliance validation without requiring manual intervention. The system automatically uploads models to validation servers, executes security tools like ClamAV and ReversingLabs, and provides deployment decisions automatically. This self-service capability reduces deployment time by eliminating manual security checking steps while maintaining thorough validation.
3Measurement precision
If comprehensive validation is performed, then detection accuracy is improved, but system complexity increases
Solution Approach 1:
The patent divides the validation system into separate validation servers with specialized functions. Primary validation servers handle compliance checking while secondary validation servers handle security scanning using tools like ClamAV and ReversingLabs. This segmentation allows each server to be optimized for its specific task, improving detection accuracy for different validation types while keeping individual server complexity manageable through functional separation.
Solution Approach 2:
The validation platform is designed as a universal system that can handle multiple validation types (compliance and security) through a unified architecture. The same validation servers can perform different validation functions depending on configuration, and the system can accommodate various model types and security tools. This multi-functionality approach improves detection accuracy across different validation scenarios while avoiding the complexity of maintaining separate specialized systems.
Data Source
AI summary
A centralized platform for validation of machine learning (ML) models for robotic process automation (RPA) before deployment is provided. The validation platform may support multiple programming languages and build platforms in a single centralized platform. The platform may allow the user to upload the model in a predefined package structure. The platform may then validate the package for deployment.