Transient Skewness

Transient skewness is the third standardized moment of the signal's amplitude distribution during the transient interval. It measures the degree of asymmetry around the mean. A positive skew indicates the tail extends toward higher amplitudes (e.g., an overshoot spike), while a negative skew indicates a tail toward lower amplitudes (e.g., a deep undershoot). Mathematically, it is computed as the expected value of the cubed deviation from the mean, normalized by the cube of the standard deviation. Unlike variance, skewness is blind to symmetric Gaussian noise, making it exceptionally useful for isolating deterministic, non-linear hardware artifacts.

Skewness in the transient envelope arises from directional non-linearities in analog components:

• Power Amplifier (PA) Asymmetry: Class AB or B amplifiers exhibit different slew rates for rising versus falling edges due to unequal charging and discharging paths in the bias network.
• Rectification Effects: Parasitic semiconductor junctions in the PA or modulator can rectify the signal, creating a DC offset that shifts the distribution's mean and induces skew.
• Power Supply Sag: The inrush current during turn-on causes a voltage drop, clipping the upper amplitude peaks and creating a negative skew in the envelope.
• Thermal Transients: Instantaneous self-heating of the transistor junction alters its gain non-linearly, often favoring one polarity of the signal swing.

A primary advantage of using skewness for fingerprinting is its theoretical insensitivity to additive white Gaussian noise (AWGN). Since the Gaussian distribution has a skewness of zero, any measured skewness in the captured signal is attributable to the transmitter's non-linear hardware, not the channel noise. This property makes skewness a highly robust feature in low signal-to-noise ratio (SNR) environments where variance-based features become contaminated. It effectively isolates the deterministic hardware signature from the stochastic noise background.

The skewness measured during the turn-on transient is often significantly more pronounced and distinct than during steady-state transmission. During steady-state operation, feedback loops (e.g., power control, bias stabilization) actively correct for non-linearities. However, during the transient, these loops are out of lock or settling, allowing the raw, uncompensated non-linearities of the hardware to dominate. This exposes the intrinsic physical asymmetry of the transistors and passives, yielding a skewness value that is a more direct and unclonable hardware fingerprint.

Transient skewness is typically used as a scalar feature in a larger fingerprinting vector:

• Sliding Window Computation: Skewness is calculated over a short, sliding time window across the transient envelope to capture its time-varying nature, producing a skewness trajectory.
• Multi-Scale Analysis: Skewness is computed on different wavelet decomposition scales to capture asymmetry at various bandwidths.
• Classifier Input: The skewness value (or trajectory) is fed into a classifier, such as a Support Vector Machine (SVM) or a shallow neural network. Because it is a single, highly discriminative number, it often provides significant separation between devices even in a low-dimensional feature space.

While both are higher-order statistics, skewness and kurtosis capture different signal properties:

• Skewness (3rd Moment): Measures asymmetry. Detects if the transient envelope has a preferred direction of deviation (e.g., more overshoot than undershoot).
• Kurtosis (4th Moment): Measures tailedness or peakedness. Detects the presence of impulsive, outlier events, such as ringing spikes or PLL glitches, regardless of their direction. Together, they form a powerful pair for characterizing the non-Gaussian, non-linear nature of the transient, with skewness providing the directional component that kurtosis lacks.

A complementary higher-order statistical measure that quantifies the peakedness and tailedness of the transient signal's amplitude probability density function. While skewness measures asymmetry, kurtosis detects the presence of impulsive, non-Gaussian artifacts caused by DAC glitches or ringing artifacts. High kurtosis values often indicate a transient dominated by rare, high-amplitude spikes, providing a distinct feature for emitter identification that is orthogonal to skewness.

The mathematical framework that generates both skewness and kurtosis as specific cases. Cumulants are higher-order statistics that are theoretically blind to Gaussian noise, making them exceptionally robust for isolating deterministic hardware non-linearities. Key applications include:

• Third-order cumulant: Directly proportional to skewness, revealing directional non-linearity.
• Fourth-order cumulant: Related to kurtosis, measuring the strength of impulsive artifacts.
• Noise suppression: Ideal for extracting the transient fingerprint from low-SNR captures.

A higher-order spectral analysis technique that reveals quadratic phase coupling within the transient signal. Unlike the power spectrum, the bispectrum preserves phase information and is zero for Gaussian processes. It effectively detects non-linear interactions, such as the mixing of PLL phase noise with the amplitude ramp, which manifest as specific skewness patterns in the time domain. This provides a frequency-domain view of the same hardware non-linearity that transient skewness measures.

The composite transient profile attributed to the power amplifier's gate or base biasing network, often the dominant physical source of amplitude asymmetry. The non-linear charging curve of the bias circuitry during the turn-on event creates a characteristic skew in the envelope's probability distribution. Analyzing this ramp signature involves:

• Slew rate measurement: The maximum dV/dt of the rising edge.
• Inflection point detection: Identifying voltage thresholds where transistor conduction modes change.
• Skewness calculation: Quantifying the asymmetry of the amplitude distribution during the ramp.

The dependence of the current transient shape on the transmitter's prior operating state, caused by thermal trapping and charge storage in semiconductor materials. This history-dependent behavior creates a dynamic skewness profile that varies with the transmitter's duty cycle and idle time. A device transmitting after a long silence will exhibit a different transient skewness than one transmitting in a rapid burst sequence, adding a temporal dimension to the fingerprint.

The collective set of statistical measures beyond second-order (variance) used to characterize the non-Gaussian nature of transient hardware artifacts. This toolkit includes:

• Skewness: Directional asymmetry of the amplitude distribution.
• Kurtosis: Peakedness and tail weight of the distribution.
• Cumulants: Noise-immune generalizations of skewness and kurtosis. These features form a robust vector for machine learning classifiers, capturing the subtle non-linearities that define a unique transient fingerprint.