Passphrase Detection Using Linguistic Context in Plain Text
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing password detection methods are ineffective in identifying passphrases containing whitespace characters and often result in high false positives due to the complex nature of passphrases, which consist of multiple words and varying word counts, making it difficult to devise a generic strategy for accurate detection.
Innovation Solution
A method and system utilizing a constituency tree-based technique combined with a pre-trained machine learning model to generate context and analyze linguistic features, such as Part-of-speech (POS) tags, to identify passphrases in plain text, incorporating a passphrase patterns database for efficient detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing password detection methods are used, then detection speed is maintained, but detection accuracy deteriorates due to inability to identify passphrases with whitespace characters
Solution Approach 1:
The passphrase detection process is segmented into distinct stages: initial scanning for potential passphrase candidates, linguistic feature extraction (POS tags, constituency trees), machine learning-based classification, and validation. This segmentation allows each stage to specialize in specific tasks, improving overall detection accuracy while managing complexity through modular processing.
Solution Approach 2:
Linguistic features serve as intermediaries between the raw passphrase text and the detection algorithm. By extracting POS tags, constituency trees, and other linguistic properties, the system creates a bridge that enables accurate identification of passphrases with whitespace characters without requiring direct complex pattern matching.
2Reliability
If generic password detection strategies are applied, then computational overhead is reduced, but false positives increase due to passphrase complexity
Solution Approach 1:
The system performs preliminary actions by pre-training machine learning models on passphrase patterns and pre-computing linguistic feature templates before actual detection. This allows the detection phase to efficiently compare against known patterns without performing complex computations in real-time, reducing false positives while maintaining productivity.
Solution Approach 2:
The detection approach changes parameters from simple character-based matching to multi-dimensional linguistic parameter analysis including POS tag sequences, constituency tree structures, and n-gram frequencies. This parameter transformation enables reliable distinction between passphrases and regular text while maintaining computational efficiency through vector-based comparisons.
3Speed
If simple detection algorithms are used, then processing speed is maintained, but ability to detect passphrases with whitespace deteriorates
Solution Approach 1:
The system replaces mechanical pattern-matching algorithms with machine learning-based classification. Instead of using rigid rules for detecting passphrases, the ML model learns complex patterns from training data, enabling it to identify whitespace-containing passphrases at speeds comparable to simple algorithms while achieving high precision.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
Nowadays, platforms are advocating the use of passphrases with the aim to provide a more secure yet memorable form of authentication as passphrases offer ease of remembering and improved adaptability to password policies without compromising usability. Existing password detection methods fail to detect passphrases due to the distinct nature of passphrases, as they involve use of multiple words, symbols, numbers, and special characters. The present disclosure provides a method and a system for detecting passphrases in plain text. The system first receives a plurality of files. Then, the system filters files based on file attributes to obtain potential files. Thereafter, a sensitivity analysis of the potential file is performed based on sensitivity indicators to obtain a sensitivity score for each potential file. Further, the system generates a set of context from text present in each potential file. Finally, the system utilizes a set of context and a sensitivity score of each potential file to identify a set of potential passphrases present in text using a pre-trained machine learning based language model.