ggerganov/kbd-audio

Features

• Acoustic Keystroke Recognition Systems - Implements a complete system for reconstructing typed text from audio recordings using waveform correlation and linguistic analysis.
• Acoustic Keystroke Recognition - Identifies which keyboard keys were pressed by analyzing the sound patterns from an audio recording.
• Acoustic Input Recognition - Identifies specific keys by comparing captured audio waveforms against trained average sound patterns using cross-correlation.
• Acoustic Profile Training - Builds prediction models by aligning and averaging audio waveforms associated with known keystrokes.
• Acoustic Keystroke Prediction - Predicts which keys were pressed by comparing captured audio waveforms against a learned training set.
• Linguistic Probability Scoring - Calculates the probability score for a string using loaded n-gram frequency data.
• N-gram Frequency Map Management - Calculates, loads, and saves n-gram probability maps for linguistic analysis and decryption.
• N-Gram Frequency Ranking - Calculates the probability of recovered text sequences by matching them against n-gram frequency maps.
• Audio Digital Signal Processing - Filters and analyzes raw audio waveforms to isolate specific sounds through threshold detection and frequency filters.
• Signal Similarity Calculation - Implements cross-correlation maps to determine how closely captured keystroke sounds match known sound profiles.
• Audio Event Detection - Identifies the exact moment of keystrokes by monitoring intensity thresholds in the audio stream.
• Sample-Threshold Detection - Identifies the precise start and end of keystrokes by monitoring intensity peaks in the audio stream.
• Acoustic Text Reconstruction - Reconstructs typed text from audio recordings using trained models and linguistic statistical frequencies.
• Cross-Correlation Analysis - Employs cross-correlation analysis to measure similarity between captured audio segments and reference sound profiles.
• Substitution Cipher Decoders - Recovers plaintext from encoded messages using n-gram frequency maps and simulated annealing algorithms.
• Side-Channel Analysis - Recovers sensitive text by processing audio signals captured from hardware during user interaction.
• Acoustic Key Detection - Locates and identifies specific key-press signals within audio waveforms using frequency filters and threshold detection.
• Acoustic Reference Profiles - Builds acoustic models by aligning and averaging multiple recordings of the same key to create a reference profile.
• Acoustic Cluster Generation - Groups audio elements into clusters using a similarity matrix and the Metropolis-Hastings algorithm.
• Similarity-Based Clustering - Implements unsupervised grouping of audio waveforms to identify repeated key-press signatures.
• Stochastic Audio Clustering - Groups similar audio waveforms into clusters using a Metropolis-Hastings stochastic sampling process.
• Frequency Filtering - Removes unwanted frequencies from waveforms using high-pass or Butterworth filters with specified cutoffs.
• Butterworth Filters - Uses Butterworth high-pass filters to remove low-frequency noise and isolate mechanical key-press transients.
• Sample Clustering - Organizes recorded keystroke sounds into groups using DBSCAN and similarity mapping to identify repeated keys.
• Interactive Audio Synthesis - Generates and manipulates sound in real time by mapping computer keyboard inputs to audio parameters.
• Keyboard-Driven Audio Generation - Produces synthesized sound in real-time by responding to computer keyboard inputs.
• Keyboard-to-Audio Translation - Translates keyboard input events into audio signals to generate sound.
• Keystroke Audio Synthesizers - Ships an audio engine that generates and manipulates synthesized sound in response to keyboard input.
• Event-Triggered Capture - Records audio from a device and stops automatically after a specified number of keystrokes are detected.
• Decryption - Uses a probabilistic simulated annealing search algorithm to break substitution ciphers by iteratively optimizing text scores.
• Offensive Security Operations - Recovers typed text by analyzing keyboard audio input.

Star history