Verification Strategy Prediction Using Obfuscated Boolean Metrics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing program verification techniques are not universally effective, and sharing programs across organizations for training ML-based strategy predictors is hindered by intellectual property concerns, leading to challenges in determining optimal verification strategies for software modules.
Innovation Solution
A method and system for privacy-preserving verification strategy prediction using Boolean relative metrics, employing obfuscated features to train a program verification strategy predictor, which extracts and obfuscates Boolean program features to minimize information leakage while predicting effective verification strategies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If program verification strategies are predicted using ML-based strategy predictors trained on actual program data, then prediction accuracy and verification effectiveness are improved, but intellectual property concerns and information leakage risks worsen
Solution Approach 1:
The patent introduces Boolean relative metrics as an intermediary representation that mediates between the original program data and the ML strategy predictor. These metrics transform program characteristics into privacy-preserving Boolean features that maintain predictive power while preventing information leakage about the actual program code and properties.
Solution Approach 2:
The patent transforms program verification data from its original form into Boolean relative metrics by changing the parameter representation. Instead of using raw program code and properties, the system converts them into Boolean features based on relative metrics, which preserves the essential information needed for prediction while eliminating sensitive details.
2Reliability
If all verification tools are run on all software modules to ensure correctness, then verification completeness is improved, but computational resource consumption and cost worsen
Solution Approach 1:
The patent applies local quality by customizing verification strategies for each software module based on its specific characteristics represented by Boolean relative metrics. Instead of applying a uniform verification approach to all modules, the system selects and orders verification tools locally for each module, optimizing resource usage while maintaining verification effectiveness.
Solution Approach 2:
The patent introduces dynamics by making verification strategies adaptive and changeable based on program characteristics. The ML-based strategy predictor dynamically selects and orders verification tools according to the specific Boolean relative metrics of each program, allowing the verification process to adapt to different software modules and their evolving characteristics.
3Reliability
If customized program verification strategies are developed for each software module, then verification effectiveness is improved, but development overhead and complexity worsen
Solution Approach 1:
The patent implements self-service by enabling software modules to effectively select their own verification strategies through the ML-based predictor. The system automatically analyzes Boolean relative metrics of each program and generates customized verification strategies without requiring manual intervention from verification experts, thereby reducing development overhead while maintaining effectiveness.
Solution Approach 2:
The patent applies universality by creating a general-purpose ML-based strategy predictor that can handle diverse software modules with different verification needs. The Boolean relative metrics framework provides a universal representation that works across different program types, and the single predictor system serves multiple verification scenarios, reducing the need for separate customization efforts for each module.
4Measurement precision
If in-house strategy predictors are developed to provide customized verification strategies, then prediction accuracy is improved, but training overhead and data requirements worsen
Solution Approach 1:
The patent uses Boolean relative metrics as an intermediary that simplifies the training process. By transforming program data into standardized Boolean features, the system creates a uniform training interface that reduces the complexity of data preparation and model training, thereby reducing training overhead while maintaining the ability to achieve high prediction accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This disclosure relates generally to method and system for a privacy preserving verification strategy prediction of an input program using Boolean relative metrics. The method extracts a plurality of Boolean Relative Metrics (BRM), and (ii) a plurality of Portfolio Driven Boolean Relative Metrics (PDBRM) from an input program based on a mode of execution for a program verification task. The method then trains a program verification strategy predictor by a strategy prediction service provider, using a plurality of obfuscated BRM corresponding to the plurality of BRM, and a plurality of obfuscated PDBRM corresponding to the plurality of PRBRM, to predict a privacy preserving program verification strategy for the program verification task, using one of a plurality of strategy prediction models in a privacy preserving Strategy Prediction (SPRED) architecture. Further the program verification strategy predictor predicts the privacy preserving program verification strategy using a plurality of Boolean feature vectors.