Confidential Data Filtering for AI Code Input Using Overlapping Hashing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The leakage of confidential information through generative AI systems poses a security risk, as program codes composed or modified using these systems may inadvertently include sensitive data, leading to confidentiality breaches.
Innovation Solution
Implementing a method that uses locality-sensitive hashing (LSH) to encode input data and compare it with secure data hash values, identifying whether the input data contains secure information by generating feature vectors and performing LSH operations on overlapping parts of the data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If generative AI systems are used to compose or modify program codes, then productivity and code generation capability are improved, but confidential information may be inadvertently included in the generated codes, leading to security risks
Solution Approach 1:
The system performs preliminary filtering of confidential information from input prompts before they are processed by the generative AI model. This prevents sensitive data from being included in the generated code, thereby maintaining both productivity and security.
Solution Approach 2:
A separate filtering system acts as an intermediary between the user input and the generative AI model. This intermediary component identifies and removes confidential information without interfering with the code generation capability of the main system.
2Reliability
If input data is filtered to remove confidential information, then data security is improved, but processing time and computational resources increase
Solution Approach 1:
The system extracts only the essential features of input data to create compact representations for comparison. This reduces the amount of data that needs to be processed during filtering, thereby maintaining high security while reducing processing time.
Solution Approach 2:
The system transforms input data into different parameter representations (hash values, feature vectors) that enable efficient comparison. By changing the form of data representation, the system achieves secure filtering without excessive processing time.
3Reliability
If comprehensive filtering of all input data is performed, then confidentiality is ensured, but device complexity and computational overhead increase
Solution Approach 1:
The system replaces complex mechanical or manual filtering processes with algorithmic approaches using hash functions and machine learning models. This substitution reduces device complexity while maintaining comprehensive confidentiality protection.
Solution Approach 2:
The system creates simplified copies or representations of confidential data (such as hash values) for comparison purposes. This allows comprehensive filtering without processing the actual complex confidential data, thereby reducing system complexity.
Data Source
AI summary
An electronic device is provided. The electronic device includes memory, including one or more storage media, storing instructions, and one or more processors communicatively coupled to the memory, wherein the instructions, when executed by the one or more processors individually or collectively, cause the electronic device to identify input data, which can be arranged in multiple lines, stored in the memory, generate a first feature vector by encoding, using an encoder, first part data corresponding to a first number of first lines among the input data, generate a second feature vector by encoding, using the encoder, second part data corresponding to the first number of second lines, the second lines at least partially overlapping the first lines, among the input data, and train the encoder such that a result of decoding the first feature vector and the second feature vector by a decoder corresponding to the encoder corresponds to the input data.