Transient Turn-On Signature

The defining characteristic is the extremely short duration, typically lasting from nanoseconds to a few microseconds. This requires high-bandwidth receivers and precise triggering mechanisms.

• Capture Window: The interval between the initial noise floor rise and steady-state modulation.
• Sampling Rate: Demands gigasamples per second (GS/s) to resolve fine-grained amplitude and phase variations.
• Triggering: Requires a high-speed, amplitude-based trigger to isolate the onset event without storing massive amounts of dead air.

The power ramp-up profile is a direct reflection of the power amplifier's (PA) non-linear charging dynamics and bias circuit stabilization. No two PAs, even from the same wafer, have identical turn-on paths.

• Rise Time: The specific duration from 10% to 90% of the final output power.
• Overshoot/Ringing: Damped oscillations caused by impedance mismatches and parasitic capacitance in the PA's matching network.
• Monotonicity: Deviations from a smooth ramp, such as small dips or inflection points, are highly discriminative hardware artifacts.

The instantaneous phase during startup is dominated by the phase-locked loop (PLL) or local oscillator's transient response as it locks to the reference frequency. This trajectory is a unique physical process.

• Settling Path: The specific spiral or damped sinusoidal path the phase takes in the I/Q plane before locking.
• Lock Time: The duration required for the phase error to fall within a specified tolerance.
• Phase Noise Burst: A temporary increase in phase noise during the aggressive frequency acquisition phase, creating a unique spectral spreading signature.

The instantaneous carrier frequency does not jump to its set point instantly; it chirps or drifts as the oscillator's control voltage stabilizes. This dynamic frequency deviation is a powerful fingerprinting feature.

• Chirp Rate: The speed of the frequency change (e.g., kHz/µs) from start-up to steady-state.
• Chirp Direction: Whether the frequency starts too high and drops, or starts too low and rises, indicating the loop filter's initial charge state.
• Spectral Splatter: The transient widening of the signal's bandwidth during the chirp, which is a direct product of the VCO's tuning sensitivity.

During the brief turn-on period, the power amplifier operates in a highly non-linear region before its bias voltages stabilize, generating a unique burst of spectral regrowth and harmonic content.

• Adjacent Channel Power (ACP) Burst: A momentary spike in out-of-band emissions that is device-specific.
• Harmonic Profile: The relative power of the 2nd, 3rd, and higher-order harmonics during the transient, which depends on the PA's semiconductor physics.
• Memory Effect Onset: The thermal and electrical memory effects of the PA manifest differently during this cold-start phase compared to steady-state operation.

For a transient signature to be a viable biometric, it must be repeatable for the same device yet unique across different devices. This is validated through statistical analysis.

• Intra-Class Variance: The minor variation in the signature across multiple turn-on events for the same device, caused by thermal noise. Must be low.
• Inter-Class Separation: The measurable distance between signature feature vectors from different devices. Must be high.
• Kolmogorov-Smirnov Test: Often used to statistically prove that the distributions of extracted transient features (e.g., fractal dimensions) are distinct for different emitters.

The turn-on transient is a brief, non-data-bearing signal segment lasting from nanoseconds to microseconds. It occurs as a transmitter's power amplifier ramps up and its local oscillator stabilizes. This period contains a unique, complex trajectory in amplitude and phase that is shaped by the specific process variations of the hardware components. Unlike the steady-state modulated signal, the transient is a direct, unfiltered expression of the transmitter's physical analog circuitry, making it a rich source for RF-DNA extraction.

Bispectrum analysis is a higher-order statistical technique critical for extracting robust features from transient signatures. It transforms the time-domain turn-on signal into a frequency domain representation that captures non-linear phase couplings unique to a transmitter's hardware. Key advantages include:

• Gaussian Noise Suppression: Inherently rejects additive white Gaussian noise.
• Phase Information Preservation: Retains signal phase, unlike the power spectrum.
• Discriminative Power: Reveals subtle non-linearities from amplifier and oscillator interactions that define a unique hardware fingerprint.

Raw transient signatures are high-dimensional vectors, making classification computationally expensive and prone to the 'curse of dimensionality'. Dimensionality reduction is a vital preprocessing step. Techniques like Principal Component Analysis (PCA) project the bispectrum or time-domain features into a lower-dimensional subspace that maximizes variance, effectively denoising the data and retaining only the most discriminative structural elements of the transient signature for efficient, real-time matching.

A transmitter's turn-on signature is not perfectly static; it drifts slowly over time due to component aging and temperature variation. A robust authentication system must implement drift compensation. This is an adaptive mechanism that continuously updates the stored reference model for a device. By using a weighted moving average of recently verified transient features, the system can track the gradual change in the fingerprint, preventing an increase in the Equal Error Rate (EER) and avoiding false rejection of a legitimate, but aged, device.

Enrolling a new device often requires capturing many transient samples, which is impractical. Contrastive learning solves this by training a deep neural network to map transient signatures into an embedding space where different captures from the same device are pulled close together, and captures from different devices are pushed apart. This allows for few-shot or even one-shot enrollment, where a device can be reliably authenticated from a single turn-on event by comparing its embedding vector to a stored template.

A core security property of transient-based authentication is its inherent replay attack resistance. An adversary cannot simply record and retransmit a valid turn-on signature because:

• The transient is a function of the physical transmitter hardware, which the attacker's own hardware cannot perfectly replicate.
• The attacker's own transmitter will imprint its own distinct power amplifier non-linearity and phase noise fingerprint onto the replayed signal, causing the combined signature to fail verification. The fingerprint is intrinsically bound to the live, physical device.