Source-linked AI summary
Near-Field Physical-Layer Authentication Under Impersonation Attacks
Hajar El Hassani, Linda Senigagliesi, Arsenia Chorti
TL;DR
The paper investigates how active attackers can impersonate transmitters in near-field PLA, where steering signatures depend on both angle and distance. It minimizes the MSE between legitimate and adversarial received signatures, deriving single- and multi-antenna conditions for exact impersonation. The main result is that near-field impersonation requires either matching Alice’s angle and distance or spanning her steering vector with Eve’s antenna array.
Problem
Near-field PLA impersonation had not been explicitly analyzed, despite near-field steering signatures depending on both angle and distance.
Method
The paper minimizes MSE between Alice’s and Eve’s received signatures and derives optimal scalar and vector precoders for single- and multi-antenna Eve.
Results
Exact impersonation requires matching Alice’s angle and distance for single-antenna Eve, or placing her steering vector in Eve’s steering-vector subspace for multi-antenna Eve.
Takeaways & Limitations
In the near field, a distance mismatch can prevent perfect impersonation even when Alice and Eve have the same AoA.
Abstract
from arXiv · showhide
This paper studies physical-layer authentication (PLA) in the near-field regime under impersonation attacks. Unlike the far-field case, where the steering vector depends only on the angle of arrival (AoA), in the near field it depends on both angle and distance. We analyze the attack by minimizing the mean-square error (MSE) between the signal received from a legitimate transmitter Alice and the signal generated by an active attacker Eve. For a single-antenna Eve, we derive the optimal scalar precoder and show that, under the standard second-order Fresnel approximation, perfect impersonation is possible only if Eve has the same angle and the same distance from Bob as Alice. We then extend the analysis to a multi-antenna Eve and derive the optimal precoding vector. In this case, perfect impersonation is possible only if Alice steering vector belongs to the subspace spanned by Eve steering vectors. Simulation results show that, in the near field, a distance difference is sufficient to prevent a successful impersonation attack even when Alice and Eve have the same AoA, and confirm the analytical results.
I. INTRODUCTION
The paper addresses near-field impersonation attacks against PLA, where authentication signatures depend on both angle and distance rather than AoA alone. It formulates single- and multi-antenna Eve attacks and derives conditions under which perfect impersonation is possible.
- Motivation: Near-field PLA is motivated by lightweight authentication needs in resource-constrained IoT networks.Conventional cryptographic authentication can add complexity, signaling overhead, and latency.
- Research gap: Unlike far-field steering vectors, near-field signatures depend jointly on angle and distance.Near-field operation arises when spherical wavefronts matter for larger arrays and higher carrier frequencies.
- Approach: The paper studies impersonation by minimizing the MSE between Alice’s received signature and Eve’s generated signature.The analysis covers both single-antenna and multi-antenna attackers.
- Contributions: For a single-antenna Eve, perfect impersonation requires the same angle and distance from Bob as Alice under the second-order Fresnel approximation.The result follows from matching the complete near-field steering signature.
- Contributions: For a multi-antenna Eve, perfect impersonation requires Alice’s steering vector to lie in the subspace spanned by Eve’s steering vectors.The paper derives the corresponding optimal precoding vector.
- Contributions: Simulations show that a distance mismatch prevents perfect impersonation even when Alice and Eve share the same AoA.The results evaluate how angle and distance affect the minimum MSE.
A. Near-field steering vector
The near-field steering vector is constructed from relative propagation distances across Bob’s receive array. Unlike in the far field, this signature depends on the transmitter’s joint angle and distance.
- Geometry: The distance from a transmitter at (r, θ) to each receive antenna is computed using the Euclidean norm.Relative path lengths are formed by referencing antenna m = 0.
- Steering vector: Under the narrowband assumption, relative path lengths determine Bob’s spatial signature through the carrier wavelength.This produces the near-field steering vector associated with transmitter position (r, θ).
- Steering vector: The near-field steering vector depends jointly on angle and distance, unlike the far-field vector.Alice, single-antenna Eve, and each multi-antenna Eve element receive distinct position-based steering vectors.
- Signal model: Alice’s transmitted signal is modeled as unit average-power narrowband data received by Bob with additive Gaussian noise.The noise variance is specified per receive antenna.
B. Single-antenna Eve model
In the single-antenna Eve model, Eve uses a complex scalar precoder to modify her transmitted signal while attempting to reproduce Alice’s received spatial signature.
- Single-antenna Eve: Eve is represented by a single antenna at pE and applies a complex scalar precoder q.The scalar precoder is the attacker’s adjustable transmission parameter in this model.
- Single-antenna Eve: The resulting received signal at Bob is determined by Eve’s position-dependent steering vector and the scalar-precoded transmission.
- Single-antenna Eve: Because Eve has one transmit antenna, her precoding can scale the signal but cannot create an independent combination of steering vectors.
C. Multi-antenna Eve model
The multi-antenna Eve model lets the attacker combine steering vectors from multiple antenna positions through a precoding vector. The resulting received signature is therefore a linear combination constrained by Eve’s steering-vector span.
- Multi-antenna Eve: Eve’s L transmit antennas each have a position-dependent steering vector that is stacked into a matrix.
- Multi-antenna Eve: Eve applies a precoding vector to combine the steering vectors associated with her transmit antennas.
- Attack objective: The attack analysis asks when Eve can make Bob’s spatial signature indistinguishable from Alice’s through precoding.The model assumes perfect knowledge of positions and array geometry and ideal precoding capabilities.
- Attack objective: Imperfect knowledge, synchronization errors, and hardware constraints would make successful impersonation more difficult for Eve.
A. Single-antenna Eve
The single-antenna attack is optimized by minimizing the MSE through a scalar precoder. Under the second-order Fresnel approximation, exact impersonation requires Eve to match Alice’s angle and distance from Bob.
- A. Single-antenna Eve: The MSE objective is reduced to minimizing the mismatch between Alice’s and Eve’s steering vectors because the noise terms do not depend on q.The resulting optimization is a real-valued quadratic handled with Wirtinger calculus.
- A. Single-antenna Eve: Equality of the steering vectors requires matching the relative path lengths across all receive antennas.This condition is analyzed using the near-field distance expression and its Fresnel expansion.
- A. Single-antenna Eve: The second-order Fresnel approximation is accurate in the radiative near-field region but may become less accurate in the extreme near field or under strong scattering.The approximation converts the relative path length into the form used for the angle-and-distance analysis.
- A. Single-antenna Eve: Exact impersonation is possible only when Eve occupies the same position as Alice, meaning the same angle and distance from Bob.Thus, under the approximation, matching angle alone is insufficient.
B. Multi-antenna Eve
For a multi-antenna Eve, the attack is optimized by choosing a precoding vector that combines Eve’s fixed steering vectors. Exact impersonation requires Alice’s steering vector to lie in their span.
- B. Multi-antenna Eve: The multi-antenna MSE optimization selects a precoding vector for Eve’s linear combination of steering vectors.The vector is obtained by minimizing the steering-vector mismatch.
- B. Multi-antenna Eve: The Moore–Penrose pseudoinverse appears in the closed-form solution for the optimal precoding vector.This solution minimizes the mismatch between Alice’s steering vector and Eve’s combined steering vector.
- B. Multi-antenna Eve: Exact impersonation is possible only if Alice’s steering vector belongs to the subspace spanned by Eve’s steering vectors.Otherwise, the optimal projection leaves a nonzero mismatch.
- B. Multi-antenna Eve: Eve cannot change the steering vectors determined by her antenna positions; precoding only controls their linear combination.The optimal precoder therefore projects Alice’s steering vector onto Eve’s generated subspace.
IV. SIMULATION RESULTS
The simulation-results section presents numerical results intended to validate the analytical derivations.
- IV. SIMULATION RESULTS: The numerical results validate the analytical derivations developed in the preceding sections.
A. Simulation setup
The simulations use a 16-element half-wavelength-spaced ULA at 2.18 GHz, with a near-field boundary of approximately 15.5 m.
- A. Simulation setup: The carrier frequency is 2.18 GHz, corresponding to a wavelength of approximately 0.138 m.
- A. Simulation setup: Bob uses a 16-element ULA with inter-element spacing λ/2 and aperture approximately 1.034 m.
- A. Simulation setup: The Fraunhofer distance is approximately 15.5 m, separating the selected near- and far-field regions.
B. Simulation figures
The simulations show that near-field impersonation accuracy depends on both distance mismatch and the multi-antenna attacker's precoding geometry. Matching Alice's angle alone is insufficient, while collocation and suitable precoding can attain the minimum MSE.
- Single-antenna Eve: With equal AoA, increasing Eve's SNR lowers MSE, but any nonzero distance mismatch leaves a larger asymptotic error that grows with the mismatch.When rE = rA, the mismatch term vanishes and MSE reduces to (1/γA + 1/γE).
- Single-antenna Eve: The MSE reaches its minimum only at ∆ = 0; distance mismatch prevents exact impersonation even when Alice and Eve share the same AoA.For larger Alice distances, the curves flatten and approach lower MSE values near the Fraunhofer distance.
- Multi-antenna Eve: In the collocated multi-antenna case, arg(q_l) = 0 satisfies the exact impersonation condition and reduces MSE to (1/γA + 1/γE).The simulation uses L = 8 Eve antennas, with Alice at (rA, θA) = (1 m, 0.4 rad).
- Multi-antenna Eve: The collocated case achieves the lowest MSE, whereas distributed antennas produce a significantly larger MSE across the considered precoding settings.The result follows because exact impersonation requires Alice's steering vector to lie in the subspace spanned by Eve's steering vectors.
V. CONCLUSION
The paper concludes that near-field physical-layer authentication imposes stricter impersonation conditions than the far-field case. Single-antenna Eve must match both Alice's angle and distance, while multi-antenna Eve must span Alice's steering vector.
- Conclusion: Near-field steering depends on both angle and distance, so exact single-antenna impersonation requires Eve to match both relative to Bob.This contrasts with far-field impersonation, where matching AoA alone can suffice.
- Conclusion: For multi-antenna Eve, exact impersonation is possible only when Alice's steering vector belongs to the subspace spanned by Eve's steering vectors.The conclusion also identifies future work on approximation accuracy and three-dimensional and MIMO extensions.