Under The Hood
Our detection engine isn't a black box. Explore the complete technical specification of every detection layer, the science behind it, and the actual implementations.
System Architecture
Real-Time Signal Processing
Neural Network Architecture
Detection Layer Specifications
Each layer implements battle-tested algorithms with documented validation notes. Click to expand and explore the science behind each detection method.
Crystal Oscillator Analysis
hologram.tsMeasures picosecond-level timing jitter from CPU crystal oscillators. VMs and emulators show zero variance due to virtualized clocks.
Exploits the fact that real quartz crystals have inherent imperfections causing measurable drift.
σ_drift = √(Σ(t_i - t̄)² / n)DOM Reflow Timing
timing-analysis.tsProfiles the cost of browser layout recalculation. Headless browsers skip this expensive operation.
Uses forced synchronous layout (offsetHeight access) to trigger reflow and measure true rendering cost.
T_reflow = t_after - t_beforeWebGL Shader Compilation
deep_identity.tsCompiles custom GLSL shaders and measures GPU driver compilation time. Each GPU model has unique timing patterns.
GPU shader compilers are highly complex and produce timing fingerprints based on hardware.
H(shader) = hash(compilation_time, error_msg)Audio DAC Latency
stealth.tsProbes the Digital-Analog Converter hardware latency through AudioContext API.
Real hardware audio systems have measurable base/output latency; headless environments return 0.
L_dac = baseLatency + outputLatencyPerformance Benchmarks
Real-world performance metrics from production deployments
Bot Detection
High-confidence detection quality
API Response
Edge-optimized global network
User Friction
Minimal impact on legitimate users
See It In Action
The live demo on our homepage runs all these detection layers against YOUR browser in real-time.