[Paper Review] Artificial-Noise Alignment for Secure Multicast using Multiple Antennas

Ashish Khist, Dongye Zhang|arXiv (Cornell University)|Nov 20, 2012

Wireless Communication Security Techniques4 references4 citations

TL;DR

This paper proposes an artificial-noise alignment scheme for secure multicasting in multi-antenna wiretap channels, where noise is aligned at legitimate receivers to preserve signal decodability while completely masking information at eavesdroppers. The scheme achieves the optimal secure degrees of freedom $ d = 1 - rac{1}{M} $ without requiring eavesdropper channel state information, relying only on known legitimate channel gains and a bound on eavesdropper SNR.

ABSTRACT

We propose an artificial-noise alignment scheme for multicasting a common-confidential message to a group of receivers. Our scheme transmits a superposition of information and noise symbols. The noise symbols are aligned at each legitimate receiver and hence the information symbols can be decoded. In contrast, the noise symbols completely mask the information symbols at the eavesdroppers. Our proposed scheme does not require the knowledge of the eavesdropper's channel gains at the transmitter for alignment, yet it achieves the best-known lower bound on the secure degrees of freedom. Our scheme is also a natural generalization of the approach of transmitting artificial noise in the null-space of the legitimate receiver's channel, previously proposed in the literature.

Motivation & Objective

• Address the scalability limitation of traditional artificial noise schemes in multicasting to multiple legitimate receivers with more receivers than transmit antennas.
• Enable secure transmission of a common confidential message to multiple single-antenna legitimate receivers using multi-antenna beamforming and artificial noise.
• Achieve optimal secure degrees of freedom $ d = 1 - rac{1}{M} $ without requiring knowledge of eavesdropper channel gains at the transmitter.
• Generalize the null-space artificial noise technique to the multicast setting by aligning noise across multiple legitimate receivers while keeping it unaligned at eavesdroppers.
• Provide a code construction that ensures reliable decoding at legitimate receivers and negligible leakage at eavesdroppers under average power constraints.

Proposed method

• Transmit a superposition of information symbols and artificial noise symbols using a precoder designed to align noise at all legitimate receivers.
• Use rationally independent channel gains to construct a codebook where noise symbols are aligned across all legitimate receivers via structured precoding.
• Employ a precoding scheme based on monomial sets $ au $ and $ ho $ to generate noise vectors aligned at legitimate receivers but not at eavesdroppers.
• Ensure that the information symbols occupy $ 1 - rac{1}{M} $ degrees of freedom by aligning noise in a subspace of dimension $ rac{1}{M} $ at legitimate users.
• Use Fano’s inequality and mutual information bounds to analyze error probability and secrecy leakage, ensuring vanishing error and leakage rates as SNR increases.
• Construct the codebook using lattice-like constellations and exploit the rational independence of channel coefficients to achieve reliable decoding and secrecy.

Experimental results

Research questions

• RQ1 Can artificial noise be aligned across multiple legitimate receivers in a multicast setting to preserve decodability while ensuring secrecy?
• RQ2 Is it possible to achieve the optimal secure degrees of freedom $ d = 1 - rac{1}{M} $ without knowledge of eavesdropper channel state information?
• RQ3 How can noise be structured so that it is aligned at legitimate receivers but not at eavesdroppers, even when eavesdropper channels are unknown?
• RQ4 What precoding and codebook design strategies ensure reliable decoding at legitimate receivers and negligible information leakage to eavesdroppers?
• RQ5 Can the degrees of freedom be maximized under average power constraints and only a bound on eavesdropper SNR?

Key findings

• The proposed artificial-noise alignment scheme achieves a secure degrees of freedom of $ d = 1 - rac{1}{M} $, matching the best-known upper bound for the compound multi-antenna wiretap channel.
• The scheme does not require knowledge of eavesdropper channel gains; it only requires a bound $ c $ on the maximum eavesdropper channel gain norm.
• The secrecy rate is bounded below by $ R ightarrow rac{1}{2} ig( (K+M)L(1- u) / (KL + L' + u) - 1 - o_P(1) ig) rac{1}{2} ext{log}_2 P $, which approaches $ rac{1}{2}(1 - rac{1}{M}) ext{log}_2 P $ as $ N o igcirc $.
• By selecting $ K = M - rac{N^{MJ_1}}{(N+1)^{MJ_1}} $, the secure degrees of freedom can be made arbitrarily close to $ 1 - rac{1}{M} $ as $ N $ increases.
• The error probability at legitimate receivers decays exponentially with SNR, ensuring reliable decoding via Fano’s inequality.
• The mutual information leakage to any eavesdropper is bounded and vanishes asymptotically, ensuring perfect secrecy in the high-SNR regime.