The 12 Real-Time Behavioral Signals.
CORROBORATE continuously analyzes 12 cognitive, neuromuscular, and temporal features across mouse, keyboard, and focus channels to verify authentic human interaction without adding user friction.
Zero PII Collection
Only differential time deltas (Δt) and relative ratios are processed. Never raw text, credentials, or intrusive video biometrics.
<20ms Edge Latency
Lightweight temporal scoring executing asynchronously in background web workers with zero main-thread UI jank.
Cognitive KLM Grounding
Enforces Card-Moran-Newell Keystroke-Level Model constraints that automated scripts and spliced bots physically violate.
Anti-Splicing Defense
Detects advanced CHIMERA attacks that splice genuine mouse donors with automated keystroke injection.
Test Biomechanical Limits in Real Time
Adjust simulated motor parameters to see how the mathematical coherence score differentiates genuine human execution from bot anomalies.
Mouse Idle During Typing Ratio
Measures whether cursor movement halts while keystrokes are being emitted.
Card-Moran-Newell KLM homing model dictates that an attending human hand cannot operate the mouse and keyboard simultaneously. High mouse velocity during typing indicates automated multi-threading or synthetic desync.
Simulated Mouse Movement during Typing (ms)
Observed temporal timing adheres to natural human neuromuscular latency bounds.
Complete 12-Signal Reference Catalog
Comprehensive mathematical formulas, cognitive explanations, and calibrated threshold boundaries.
Mouse Idle During Typing Ratio
Measures whether cursor movement halts while keystrokes are being emitted.
\text{MIDT} = 1 - \frac{\sum \Delta t_{\text{mouse-move}}}{\text{TypingSpan}}Concurrent Operation Cost
Penalizes zero-latency transitions between distinct modality channels.
\text{COC} = \exp\left(-\frac{(\Delta t_{\text{switch}} - \mu_{\text{homing}})^2}{2\sigma^2}\right)Parallel Pointer-Key Cross-Rate
Instantaneous cross-channel event density within micro-windows.
\text{PPX} = \frac{\text{Count}(M_t \cap K_t \mid \Delta t < 20\text{ms})}{\text{Total Events}}Pointer Precedes Focus Latency
Checks if cursor trajectory physically leads focus acquisition.
\text{PPF} = \text{clamp}\left(1 - \frac{t_{\text{focus}} - t_{\text{last-move}} - 400}{800}, 0, 1\right)Keystroke Onset Lag
Latency between clicking into an input field and the first character keydown.
\text{KOL} = \exp\left(-0.5 \left(\frac{\Delta t - 300}{220}\right)^2\right)Scroll-to-Interact Latency
Settling time between viewport deceleration and subsequent interaction.
\text{STI} = \text{Plikelihood}(\Delta t_{\text{scroll-end} \to \text{click}})Action-Before-Commit Hesitation
Pre-submission micro-pauses reflecting task confirmation.
\text{ABC} = \sigma(\Delta t_{\text{hover-submit} \to \text{click}})Click-to-Input Alignment
Spatial-temporal bounding of clicks relative to input bounding boxes.
\text{CIA} = \text{PDF}_{\text{Gaussian}}(\text{Offset}_{\text{center}})Inter-Channel Pace Correlation
Correlation between mouse velocity cadence and typing speed bursts.
\text{IPC} = \text{Corr}(\mathbf{v}_{\text{mouse}}, \mathbf{r}_{\text{keys}})Rhythm Variance Consistency
Stability of micro-pause distributions across form episodes.
\text{RVC} = \frac{\text{MAD}(\Delta t_{\text{flight}})}{\text{Mean}(\Delta t_{\text{flight}})}Field-Episode Completeness
Validates that entry, keystroke modification, and blur form a complete episode.
\text{FEC} = \prod_{f \in \text{Fields}} \mathbb{I}(E_f \prec K_f \prec B_f)Form-Traversing Interleaving Pattern
Logical progression across DOM elements versus random injection.
\text{FTI} = \text{LevenshteinSimilarity}(\text{ObservedPath}, \text{DOMGraph})