Plaintext Passphrase Detection with Contextual Language Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing password detection methods are inadequate for identifying passphrases in plaintext due to their distinct nature, which involves multiple words, symbols, and whitespace characters, leading to challenges in memorization and security risks.
Innovation Solution
A method and system using a constituency tree-based technique and a pre-trained machine learning language model to identify passphrases in plaintext by analyzing file sensitivity, generating context, and utilizing POS tags to detect potential passphrases, with a fine-tuning mechanism for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing password detection methods are used, then detection simplicity is maintained, but detection accuracy for passphrases deteriorates due to their distinct nature with multiple words and whitespace characters
Solution Approach 1:
The passphrase detection process is segmented into multiple specialized stages: file filtering based on attributes, sensitivity analysis to score potential files, constituency tree generation for contextual structure, and machine learning-based passphrase identification. Each stage handles a specific aspect of the detection task, improving overall accuracy while managing complexity through modular design.
Solution Approach 2:
Constituency trees serve as an intermediary structure that bridges the gap between raw text and passphrase detection. These trees provide contextual information and syntactic structure that help the machine learning model distinguish passphrases from regular text, enabling accurate detection without requiring the model to process raw text directly.
2Ease of operation
If plaintext storage is used to reduce memorization burden, then ease of operation is improved, but security deteriorates due to exposure risks
Solution Approach 1:
The system performs preliminary detection of passphrases in plaintext files before security breaches can occur. By proactively identifying and flagging stored passphrases, the system enables users to secure or remove sensitive credentials before they can be exploited by attackers who might gain access to the device or files.
Solution Approach 2:
The system provides feedback to users about detected passphrases, including their locations and security risks. This feedback loop enables users to make informed decisions about securing their credentials, balancing the convenience of plaintext storage with security considerations.
3Reliability
If comprehensive file analysis is performed to improve detection coverage, then detection completeness is improved, but computational overhead increases
Solution Approach 1:
File filtering based on attributes and sensitivity analysis are performed as preliminary steps before comprehensive passphrase detection. This preliminary processing identifies and prioritizes files that are most likely to contain passphrases, allowing the system to focus computational resources on high-value targets rather than analyzing every file equally.
Solution Approach 2:
The system applies different levels of analysis to different files based on their sensitivity scores and attributes. High-sensitivity files receive comprehensive analysis with constituency tree generation and machine learning detection, while low-sensitivity files receive minimal or no analysis, optimizing the balance between detection coverage and computational overhead.
Data Source
AI summary
Nowadays, platforms are advocating use of passphrases with aim to provide more secure yet memorable form of authentication as passphrase offer ease of remembering, and improved adaptability to password policies without compromising usability. Existing password detection methods fail to detect passphrases due to distinct nature of passphrases, as it involves use of multiple words, symbols, numbers, and special characters. Present disclosure provides method and system for detecting passphrases in plain text. The system first receives plurality of files. Then, system filters files based on file attributes to obtain potential files. Thereafter, sensitivity analysis of potential file is performed based on sensitivity indicators to obtain sensitivity score for potential file. Further, system generates set of context from text present in each potential file. Finally, system utilizes set of context and sensitivity score of potential file to identify set of potential passphrases present in text using pre-trained machine learning based language model.