[Paper Review] Secure Communications over Fading Channels

Yingbin Liang, H. Vincent Poor|ArXiv.org|Aug 20, 2007

Wireless Communication Security TechniquesEngineering11 references16 citations

TL;DR

This paper establishes the secrecy capacity for fading wire-tap channels with full channel state information (CSI) at both transmitter and receiver. It proposes an optimal power allocation strategy that dynamically allocates more power when the legitimate channel is strong and the eavesdropper channel is weak, significantly outperforming uniform power allocation, especially under Rayleigh fading with CSI knowledge.

ABSTRACT

The fading wire-tap channel is investigated, where the source-to-destination channel and the source-to-wire-tapper channel are corrupted by multiplicative fading gain coefficients in addition to additive Gaussian noise terms. The channel state information is assumed to be known at both the transmitter and the receiver. The parallel wire-tap channel with independent subchannels is first studied, which serves as an information-theoretic model for the fading wire-tap channel. The secrecy capacity of the parallel wire-tap channel is established. This result is then specialized to give the secrecy capacity of the fading wire-tap channel, which is achieved with the source node dynamically changing the power allocation according to the channel state realization. An optimal source power allocation is obtained to achieve the secrecy capacity.

Motivation & Objective

• To establish the secrecy capacity of a parallel wire-tap channel with independent subchannels, where each subchannel is a general broadcast channel without a degradedness assumption.
• To specialize the parallel wire-tap channel result to the fading wire-tap channel, modeling time-varying fading as independent subchannels.
• To derive the optimal power allocation policy at the source that maximizes secrecy rate under an average power constraint and full CSI.
• To demonstrate that the optimal power allocation differs from water-filling and is crucial for achieving high secrecy capacity in fading environments.
• To compare the performance of optimal power allocation with uniform power allocation under different fading distributions.

Proposed method

• Model the fading wire-tap channel as a parallel wire-tap channel with L independent subchannels, each corresponding to a fading state realization.
• Establish the secrecy capacity of the general parallel wire-tap channel using information-theoretic techniques, including the use of auxiliary random variables and mutual information constraints.
• Apply the result to the fading wire-tap channel by treating each fading state as a subchannel, with the source knowing both the main and eavesdropper channel gains.
• Derive the optimal power allocation P*(h̲) that maximizes the ergodic secrecy rate, using the Kuhn-Tucker conditions on the concave objective function involving log-likelihood ratios of SNR gains.
• Prove that subchannels where the eavesdropper's channel is stronger than the destination's (i.e., h̲ ∈ Aᶜ) should be inactive (power = 0), reducing the problem to only the subchannels where the destination's channel is better.
• Use numerical simulations to compare the secrecy capacity under optimal power allocation with that under uniform power allocation across fading states.

Experimental results

Research questions

• RQ1 What is the secrecy capacity of a parallel wire-tap channel with non-degraded, independent subchannels?
• RQ2 How does the secrecy capacity of a fading wire-tap channel depend on channel state information at the transmitter?
• RQ3 What is the optimal power allocation policy that maximizes the ergodic secrecy rate in a fading wire-tap channel with full CSI?
• RQ4 How does the optimal power allocation differ from the water-filling strategy used in non-secretive fading channels?
• RQ5 Under what fading distributions does uniform power allocation perform close to optimal in terms of secrecy rate?

Key findings

• The secrecy capacity of the fading wire-tap channel is achieved by allocating more power to channel states where the source-to-destination channel gain |h₁|² is large and the source-to-wire-tapper channel gain |h₂|² is small.
• The optimal power allocation policy is derived using the Kuhn-Tucker conditions and is not equivalent to the water-filling strategy used in non-secretive fading channels.
• Subchannels where the wire-tapper's channel is stronger than the destination's (|h₂|² > |h₁|²) should be inactive, as they do not contribute to secrecy capacity.
• Numerical results show that optimal power allocation significantly outperforms uniform power allocation in Rayleigh fading, especially at moderate SNRs.
• For uniformly distributed fading gains over discrete points, uniform power allocation approaches the secrecy capacity at high SNR, indicating that CSI gain is less critical under certain distributions.
• The ergodic secrecy capacity is given by the expectation over fading states of the difference between the mutual information at the destination and the equivocation at the wire-tapper, maximized over power allocation.