Password Discovery Using Generative Adversarial Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing password discovery techniques, such as brute-force lookups and machine learning algorithms, are inefficient and ineffective in identifying both known and unknown passwords in large data sets, due to computational expense, reliance on accurate and exhaustive password datasets, and difficulty in training accurate machine learning models.
Innovation Solution
The use of specialized generative adversarial networks (GANs) trained with dictionary data sets selected or generated using specific rules to model human password generation patterns, allowing for the generation of massive potential password datasets that can be compared against large corpuses of data to identify unknown passwords.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If brute-force lookup techniques are used to search for known passwords in large data sets, then previously-known passwords can be identified, but the computational expense and time required become intractable when data sets contain billions of documents
Solution Approach 1:
The patent applies preliminary action by pre-processing the training data to extract and store password patterns, character sequences, and statistical features before the actual search. This preprocessing creates ready-to-use pattern databases that enable rapid comparison during the search phase, avoiding the need to analyze raw text data in real-time and thus resolving the contradiction between thorough password detection and search efficiency
Solution Approach 2:
The patent replaces the mechanical brute-force comparison approach with a pattern-matching system that uses statistical models and learned password patterns. Instead of comparing every possible password candidate against the data set, the system uses trained models to identify likely password locations through pattern recognition, substituting computational mechanics with intelligent pattern-based detection that scales efficiently to billions of documents
2Device complexity
If machine learning classification algorithms trained on small random samples of known passwords are used, then the system can operate with limited training data, but the algorithms return a high number of false negatives and false positives
Solution Approach 1:
The patent applies preliminary action by implementing extensive data augmentation and preprocessing steps before training the machine learning models. This includes generating synthetic password samples, extracting diverse password patterns from available data, and creating balanced training sets with various password types and lengths. These preliminary preparations ensure that even with limited initial training data, the models receive comprehensive and representative training samples, thereby improving accuracy without requiring excessively large training data sets
Solution Approach 2:
The patent implements self-service through automated feedback loops where the system uses its own outputs to improve its training data. Discovered passwords and pattern matches are fed back into the training pipeline to generate additional training samples, allowing the system to continuously improve its accuracy while operating with initially limited training data. This self-reinforcing mechanism resolves the contradiction by enabling the system to bootstrap its performance from small samples
3Measurement precision
If the dictionary used for brute-force detection is enlarged to improve accuracy of detecting novel passwords, then the accuracy increases, but the computational complexity becomes a function of both the corpus size and dictionary size, making the task intractable
Solution Approach 1:
The patent applies segmentation by dividing the large dictionary into multiple smaller, organized subsets based on password characteristics such as length, character type, and pattern categories. The system processes these segmented dictionary portions separately and combines results, which reduces the computational complexity of comparing against the entire dictionary at once while maintaining comprehensive coverage for detecting novel passwords
Solution Approach 2:
The patent implements local quality by applying different dictionary subsets and matching strategies to different regions or contexts within the data corpus. Instead of uniformly applying the entire dictionary to all data, the system adapts the dictionary selection and comparison methods to local characteristics of the data being analyzed, improving detection accuracy for novel passwords while reducing unnecessary computational complexity in regions where full dictionary scanning is not needed
Data Source
AI summary
Systems and methods for password discovery are provided. A system receives a first password data set comprising known passwords and applies a rule-set to the first data set to generate a second password data set comprising passwords that are believed to be likely to be human-generated. The system trains a generative adversarial network, for generating predicted passwords, using the second data set, for example by incentivizing the GAN to favor passwords in the second data set. The system applies the generative adversarial network to generate a third password data set comprising predicted passwords. The system compares the third password data set to a data corpus to identify a string in the data corpus determined to match one of the predicted passwords in the first plurality of predicted passwords. The identified string may thus be identified as a previously undiscovered password, which may be applied to unlock password-protected systems and/or to further improve password discovery systems.