I/Q Imbalance

A difference in amplification between the In-Phase (I) and Quadrature (Q) branches of the modulator. This amplitude discrepancy causes the constellation diagram to stretch or compress along one axis, creating an elliptical rather than circular distribution of symbol points.

• Measured in decibels (dB) as the ratio of I to Q gain
• Typical values range from 0.1 dB to 2 dB in commercial hardware
• Results in mirror-frequency interference in the transmitted spectrum
• Creates a stable, measurable feature for device identification

The deviation from the ideal 90-degree phase offset between the I and Q local oscillator signals. Instead of being perfectly orthogonal, the two branches exhibit a slight phase skew, causing cross-talk between the I and Q channels.

• Measured in degrees of deviation from orthogonality
• Typical errors range from 0.5° to 5° in practical transmitters
• Causes constellation points to rotate and shear
• Phase error is highly temperature-dependent, adding a temporal dimension to the fingerprint

A constant voltage bias present in the I or Q signal path that shifts the entire constellation away from the origin. This carrier leakage produces an unmodulated tone at the center frequency of the transmitted signal.

• Arises from component mismatches in differential circuits
• Creates a distinctive spike at the carrier frequency in the power spectrum
• Magnitude is typically -25 dBc to -40 dBc relative to the modulated signal
• Highly stable over time, making it a reliable long-term fingerprint feature

Unlike static gain and phase errors, frequency-dependent I/Q imbalance varies across the signal bandwidth. It is caused by mismatched low-pass filters in the I and Q reconstruction paths and group delay differences between the two branches.

• Manifests as frequency-selective image rejection degradation
• Requires wideband analysis to fully characterize
• Creates a richer, higher-dimensional fingerprint than static imbalance alone
• Particularly pronounced in direct-conversion transmitters with wide modulation bandwidths

The primary metric quantifying the severity of I/Q imbalance. IRR measures the power ratio between the desired signal and the unwanted mirror-frequency image created by gain and phase mismatches.

• Calculated as: IRR = 10 log₁₀(P_desired / P_image)
• A perfectly balanced modulator has infinite IRR
• Typical commercial hardware achieves 25–40 dB IRR
• IRR degrades with temperature drift and component aging, creating a slowly evolving fingerprint

The combined visual manifestation of all I/Q imbalance components on the constellation diagram. Each transmitter produces a unique warping pattern—a systematic displacement of received symbols from their ideal reference positions.

• Gain mismatch: elliptical stretching along I or Q axis
• Phase error: rhomboidal shear of the constellation
• DC offset: global translation away from origin
• This composite warping pattern serves as a visual fingerprint and is the basis for many feature extraction algorithms in SEI systems

A foundational metric quantifying the deviation of received constellation points from their ideal reference positions. EVM captures the aggregate effect of all hardware impairments, including I/Q imbalance, phase noise, and amplifier non-linearity. It serves as a primary feature input for many deep learning-based fingerprinting models, providing a compact representation of transmitter-specific distortion. A higher EVM generally indicates a more pronounced and potentially more identifiable hardware signature.

The unique spectral broadening signature caused by short-term random frequency fluctuations in a transmitter's local oscillator. Unlike I/Q imbalance which is a static offset, phase noise is a dynamic, random process that creates a distinctive 'skirt' around the carrier frequency. This fingerprint is highly individual to each oscillator and is notoriously difficult to clone, making it a robust feature for high-security device authentication even when other impairments are subtle.

The distinctive distortion pattern introduced when a transmitter's power amplifier operates near its saturation point. Characterized by AM/AM conversion (amplitude-dependent gain compression) and AM/PM conversion (amplitude-dependent phase shift), this non-linear behavior creates unique spectral regrowth and in-band distortion. Neural networks can learn to isolate this fingerprint from other channel effects, using it as a highly discriminative identifier for specific transmitter hardware models and individual units.

A neural network architecture that directly processes in-phase and quadrature (I/Q) samples as complex numbers, preserving the critical phase and magnitude relationships. Unlike real-valued networks that treat I and Q as separate channels, CVNNs inherently model the rotational nature of I/Q imbalance and phase errors. This architectural choice often yields superior fingerprinting accuracy because it respects the underlying physics of the signal, learning more robust and interpretable representations of hardware impairments.

Techniques designed to extract transmitter-specific features that remain stable and discriminative despite varying multipath propagation and channel impairments. The central challenge is that the wireless channel can mask or distort the subtle hardware fingerprint. Solutions include domain adversarial training, which forces a feature extractor to confuse a channel classifier, and contrastive learning, which pulls same-device signals from different channels together in embedding space while pushing different devices apart.

A security primitive that derives a unique, unclonable cryptographic identity from the inherent, random manufacturing variations in a device's RF analog front-end. I/Q imbalance, along with oscillator phase noise and amplifier non-linearity, is a primary source of entropy for RF PUFs. This approach binds a device's identity directly to its physical hardware, making it fundamentally resistant to spoofing attacks that rely on stealing software-based credentials or MAC addresses.