I/Q Distortion Signature

The signature must be sufficiently distinct across a population of devices to enable reliable identification. This uniqueness arises from the random, uncorrelated nature of manufacturing variances in analog components like mixers, filters, and data converters.

• Statistical Independence: The specific combination of I/Q gain imbalance, quadrature skew, and DC offset forms a high-dimensional vector that is statistically unlikely to repeat.
• Physical Unclonability: The signature is an emergent property of the physical hardware, making it impossible to replicate in a different device, even with identical make and model.

For a signature to be a viable identifier, it must remain consistent and repeatable over short time intervals under fixed environmental conditions. The distortion pattern measured today must correlate strongly with the pattern measured moments later.

• Short-Term Consistency: The signature should exhibit minimal variance when measured across multiple consecutive transmissions.
• Measurement Confidence: High repeatability allows for the establishment of a tight statistical boundary around the signature, reducing false rejection rates in authentication systems.

While stable in the short term, the signature exhibits predictable drift over longer periods due to environmental factors. This is a critical characteristic for practical deployment.

• Thermal Dependence: Component values, and thus the I/Q imbalance, change with temperature. A signature is often characterized by its thermal drift profile.
• Aging Effects: Over months and years, component degradation causes a slow, secular drift in the signature, requiring adaptive enrollment algorithms to track the legitimate device's evolving fingerprint.

The signature is not a single number but a complex geometric deformation of the ideal constellation. Its morphology is described by a vector of interacting parameters.

• Constellation Warping: The ideal square or circular grid is distorted into a parallelogram, trapezoid, or other non-uniform shape due to the combined effect of gain and phase errors.
• Point Cloud Statistics: Each symbol's cluster of points is characterized by its centroid offset, ellipticity, and tilt angle, providing a rich feature set for machine learning classifiers.

The distortion signature is often not constant across all operating conditions. It can vary as a function of the transmitted signal itself, revealing deeper hardware non-idealities.

• Power-Dependent Impairment: The level of I/Q imbalance and DC offset can change with the transmitter's output power level, creating a signature profile across a power range.
• Frequency-Dependent Response: The analog front-end's gain and phase flatness varies across the channel bandwidth, meaning the distortion signature can differ for signals at different carrier frequencies.

A practical signature must be extractable and verifiable even after the signal has passed through a non-ideal wireless channel with multipath fading and noise.

• Channel-Resilient Features: The core geometric distortions (e.g., the shape of the warped constellation) are deterministic and imposed at the transmitter, making them separable from the random, additive effects of the channel.
• Contrastive Learning: Modern extraction techniques use deep learning models trained to ignore channel-specific variations and focus on the invariant, hardware-specific distortion pattern.

A hardware impairment in direct-conversion transmitters where the in-phase (I) and quadrature (Q) signal paths exhibit mismatched amplitude or phase. This mismatch creates a unique, identifiable distortion in the constellation diagram, forming the foundation of the I/Q distortion signature.

• Gain imbalance: Unequal amplitude between I and Q paths
• Phase imbalance: Deviation from the ideal 90-degree separation
• Results in constellation warping from a square to a parallelogram

The geometric deformation of an ideal constellation diagram into a non-uniform shape caused by the combined effects of I/Q gain and phase imbalances. This warping is deterministic and repeatable for a given transmitter.

• Manifests as a parallelogram or trapezoid instead of a square
• The specific warping angle and ratio form a unique device signature
• Quantified by quadrature skew and I/Q gain ratio measurements

The displacement of the constellation diagram's center from the ideal (0,0) coordinate, primarily caused by DC offset and local oscillator leakage in the transmitter's analog stages. This offset is a stable, measurable component of the distortion signature.

• Caused by carrier feedthrough in the mixer stage
• DAC offset errors contribute static displacement
• The offset vector's magnitude and angle are device-specific

The statistical dispersion and shape characteristics of measured signal points around each ideal constellation locus. Beyond simple Error Vector Magnitude (EVM), the cloud's ellipticity, tilt angle, and higher-order moments form a rich feature space for emitter identification.

• Ellipticity reveals the ratio of I/Q gain to phase imbalance
• Tilt angle provides a sensitive phase imbalance measure
• Kurtosis and skewness capture non-Gaussian noise signatures

The degree to which a transmitter's I/Q impairment signature remains constant over short time intervals under fixed environmental conditions. Signature stability is a critical requirement for reliable physical layer authentication.

• Must remain consistent across multiple transmission bursts
• Short-term stability enables reliable device re-identification
• Contrasts with distortion drift caused by temperature and aging

A digital signal processing technique that dynamically estimates and compensates for time-varying I/Q imbalance and DC offset using feedback loops or blind estimation algorithms. While correction improves modulation fidelity, residual uncorrected errors often remain and can serve as a persistent fingerprint.

• Blind estimation uses statistical properties of the received signal
• Imperfect correction leaves behind a unique residual signature
• The correction algorithm's limitations become part of the fingerprint