The Wiretapped Diamond-Relay Channel

Si-Hyeon Lee, Ashish Khisti

In the push toward 5G and beyond, the Cloud Radio-Access Network (C-RAN) has emerged as a frontrunner architecture. In a C-RAN, simple base stations act as relays, connected via high-speed backhaul links to a central "cloud" that handles heavy-duty processing. However, the security of these backhaul links is often overlooked. If an eavesdropper taps into these links while also listening to wireless transmissions, the system's integrity is at risk.

A recent paper by Si-Hyeon Lee and Ashish Khisti, titled "The Wiretapped Diamond-Relay Channel," provides a definitive information-theoretic answer to this problem.

System Model

The authors model this scenario as a Diamond-Relay Channel consisting of two hops. In the first hop, a source connects to $M$ relays via orthogonal links. In the second hop, these relays transmit to a destination over a wireless Multiple-Access Channel (MAC).

The "Wiretapped" variant assumes an eavesdropper observes the wireless transmissions and taps into $W$ of the $M$ source-relay links. Crucially, the legitimate nodes know that $W$ links are compromised but do not know which specific ones. The objective is to maximize the secure degrees of freedom (s.d.o.f.), ensuring the eavesdropper learns nothing about the message.

The Wiretapped Diamond-Relay Channel model.

Methodology

The paper establishes the optimal s.d.o.f. through a two-pronged approach:

1. Achievability

To handle the uncertainty of which links are tapped, the authors developed schemes combining Secure Network Coding (SNC) and Interference Alignment:

  • SNC over Real Numbers: The source introduces "fictitious message symbols" (randomness) that mask the real data on the backhaul links.
  • Interference Alignment: At the relays, these fictitious symbols and artificial noise are precoded so they "align" (cancel out) at the legitimate destination but remain distinct and masking at the eavesdropper.

2. Converse

The optimality is proven using a genie-aided approach. The authors show that even if a "genie" informed the source exactly which links were wiretapped, the maximum achievable s.d.o.f. would not increase. This proves their "blind" scheme is optimal.

Key Results

The optimal s.d.o.f. ($d_s$) is defined as a function of the backhaul link capacity ($\alpha$). The results highlight a threshold behavior based on the number of secure links ($N = M - W$):

  • For $N = 1$ (One secure link): $$d_s = \min \left( \alpha, \frac{M-1}{M} \right)$$
  • For $N \ge 2$ (Two or more secure links): $$d_s = \min \left( N\alpha, \frac{N\alpha + M - 1}{M+1}, 1 \right)$$
Secure degrees of freedom (s.d.o.f.) vs. backhaul capacity alpha.

As shown in the figure, if at least two secure backhaul links exist ($N \ge 2$), sophisticated alignment techniques can completely nullify the eavesdropper's advantage on the wireless hop as backhaul capacity ($\alpha$) increases, eventually reaching a full s.d.o.f. of 1.

Conclusion and Future Work

This work provides a definitive information-theoretic bound for securing C-RAN architectures against multi-layer eavesdropping. While the study focuses on the high-SNR limit (degrees of freedom), it sets the stage for future research into secrecy capacity for low-power environments and more complex network topologies.

Summary updated

BibTeX
@article{lee2018wiretapped,
  author = {Si-Hyeon Lee and Ashish Khisti},
  title = {{The Wiretapped Diamond-Relay Channel}},
  journal = {IEEE Trans. Information Theory},
  year = {2018},
  doi = {10.1109/tit.2018.2866146},
}
Figure