Mimicry Attack Query Generation for Injection Prevention Testing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing runtime query injection prevention models are vulnerable to mimicry-based attacks that disguise data exfiltration by modifying benign queries without altering syntax or introducing additional features, leading to unwarranted information disclosure.
Innovation Solution
A mimicry-based attack generation system that generates exploration and exploitation queries to test and update injection prevention models, using exploration models to mutate benign queries and exploitation models to introduce protected data identifiers, thereby identifying and updating the models to prevent data exfiltration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If injection prevention models use syntax-based or feature-based approaches to detect malicious queries, then they can identify obvious attacks, but they fail to detect mimicry-based attacks that preserve benign query syntax and features
Solution Approach 1:
The system dynamically adapts the injection prevention model through iterative training. Exploration queries that successfully evade detection are added to the training set, transforming static syntax/feature-based rules into a dynamic system that continuously improves detection capabilities against evolving mimicry attacks
Solution Approach 2:
The system performs preliminary exploration attacks during the training phase to identify vulnerabilities in the injection prevention model before deployment. By pre-generating exploration queries and testing them against the model, the system proactively strengthens detection capabilities before facing real attacks
2Reliability
If exploration queries are generated to test injection prevention models, then the models can be strengthened against mimicry attacks, but the testing process requires significant computational resources and time
Solution Approach 1:
The system uses a two-stage approach where exploration queries first test syntax-based detection, then exploitation queries test feature-based detection only on queries that pass the first stage. This partial action through staged testing reduces overall computational overhead while maintaining comprehensive evaluation
Solution Approach 2:
The attack generation process is segmented into distinct exploration and exploitation stages. Exploration queries focus on syntax manipulation while exploitation queries focus on feature manipulation, dividing the complex testing task into manageable segments that can be processed efficiently
Data Source
AI summary
A method implements mimicry-based attack generation. The method may include applying an exploration model to a first query of a set of benign queries to generate an exploration query and applying an injection prevention model to the exploration query to generate an exploration result. The method may further include updating the set of benign queries to include the exploration query when the exploration result indicates the exploration query was accepted and applying an exploitation model to a second query of the set of benign queries to generate an exploitation query comprising a protected data identifier. The method may include applying the injection prevention model to the exploitation query to generate an exploitation result and storing the exploitation query as an exfiltration query when the exploitation result comprises protected data accessed with the protected data identifier.