Deceptive Infusion of Data for AI-ML Privacy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Proprietary scientific data owners face challenges in sharing data with AI/ML researchers due to concerns about reverse-engineering and disclosure of sensitive information, as conventional privacy-preserving methods do not effectively protect the underlying governing laws of systems while allowing AI/ML processing.
Innovation Solution
The DIOD methodology decomposes raw data into fundamental and inference metadata, using non-invertible mathematical transformations and concealment operators to obfuscate the fundamental metadata, preserving AI/ML correlations while concealing system identity, employing reduced order modeling and deception kernels to ensure data protection and maintain AI/ML performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If proprietary scientific data is shared publicly with AI/ML researchers, then data accessibility and reusability are improved, but sensitive information and system identity may be compromised through reverse engineering
Solution Approach 1:
The patent segments raw scientific data into two distinct metadata components: fundamental metadata (containing system identity and governing laws) and inference metadata (containing AI/ML processing information). This segmentation allows selective sharing where only inference metadata is made accessible to researchers, while fundamental metadata remains protected, thus enabling data accessibility without exposing sensitive system information to reverse engineering
Solution Approach 2:
The patent introduces an intermediary obfuscation layer that transforms fundamental metadata into obfuscated fundamental metadata using deception kernels and concealment operators. This intermediary transformation preserves the utility of data for AI/ML processing while concealing the original system identity, acting as a mediator between data sharing goals and security requirements
2Object-affected harmful factors
If data is sanitized to remove sensitive information, then confidentiality is improved, but AI/ML processing capability may be degraded due to loss of essential patterns
Solution Approach 1:
The patent applies parameter changes by transforming fundamental metadata through deception kernels that modify its mathematical representation while preserving the underlying relationships and patterns necessary for AI/ML processing. The obfuscation transforms parameters such as system identity descriptors while maintaining the structural relationships that enable machine learning inference, thus protecting confidentiality without degrading processing capability
3Object-affected harmful factors
If conventional privacy-preserving methods are used, then some sensitive information is protected, but the underlying governing laws of systems remain vulnerable to disclosure
Solution Approach 1:
The patent extracts the underlying governing laws and system identity information into a separate fundamental metadata category that is then independently obfuscated using deception kernels. This extraction ensures that conventional privacy-preserving methods are enhanced by specifically targeting and protecting the governing laws through mathematical transformations, preventing their disclosure while still allowing sanitized data to be shared
Data Source
AI summary
Systems, apparatuses, and methods for deceptive infusion and obfuscation of data are disclosed. An apparatus including a communication terminal and a processing circuitry. The communication terminal is configured to transmit information to an artificial intelligence engine. The processing circuitry is configured to decompose raw data into fundamental metadata and inference metadata. The processing circuitry is also configured to generate one or more concealment operators and generate a deception kernel responsive to the inference metadata, the one or more concealment operators, and/or the fundamental metadata. The processing circuitry is configured to obfuscate the fundamental metadata responsive to the one or more concealment operators and the deception kernel, and provide the obfuscated fundamental metadata and the inference metadata to the artificial intelligence engine for processing.