Randomization for Security in Half-Duplex Two-Way Gaussian Channels

This paper develops a new physical layer framework for secure two-way wireless communication in the presence of a passive eavesdropper, i.e., Eve. Our approach achieves perfect information theoretic secrecy via a novel randomized scheduling and power allocation scheme. The key idea is to allow Alice and Bob to send symbols at random time instants. While Alice will be able to determine the symbols transmitted by Bob, Eve will suffer from ambiguity regarding the source of any particular symbol. This desirable ambiguity is enhanced, in our approach, by randomizing the transmit power level. Our theoretical analysis, in a 2-D geometry, reveals the ability of the proposed approach to achieve relatively high secure data rates under mild conditions on the spatial location of Eve. These theoretical claims are then validated by experimental results using IEEE 802.15.4-enabled sensor boards in different configurations, motivated by the spatial characteristics of Wireless Body Area Networks (WBAN).

💡 Research Summary

The paper introduces a novel physical‑layer security scheme for half‑duplex two‑way Gaussian channels in the presence of a passive eavesdropper (Eve). Unlike traditional approaches that rely on channel state information (CSI) or artificial noise injection, the authors exploit randomness in both transmission timing and power level to create ambiguity about the source of each received symbol. Alice and Bob each decide independently, according to a Bernoulli process, whether to transmit in a given time slot. When a transmission occurs, the node selects its transmit power from a pre‑defined probability distribution. Because Alice and Bob share the random seed and the power‑level distribution beforehand, each can perfectly identify the other’s symbols despite the half‑duplex constraint. Eve, however, observes only the superposition of signals and cannot reliably infer which node generated a particular symbol; the probability of correct source identification converges to 0.5, which corresponds to perfect information‑theoretic secrecy.

The authors develop a rigorous 2‑D geometric model to quantify secrecy. Alice and Bob are placed at fixed positions, while Eve’s location is parameterized by distance (d) and angle (\theta). For each transmission event the paper derives the conditional entropy of the source given Eve’s observation and the mutual information between the legitimate parties and Eve. Using Markov’s inequality and algebraic bounds, a lower bound on the achievable secure rate is obtained. The analysis shows that, provided Eve is not too close to the line segment joining Alice and Bob (or is sufficiently angularly offset), the random scheduling and power allocation dramatically increase Eve’s error probability while preserving the average signal‑to‑noise ratio (SNR) at the legitimate receiver. Consequently, the scheme can sustain relatively high secure data rates even under modest power constraints typical of low‑power sensor networks.

To validate the theory, the authors implement the protocol on IEEE 802.15.4‑compatible TI CC2420 sensor boards (TelosB). Experiments emulate a Wireless Body Area Network (WBAN) scenario: Alice and Bob are placed on opposite sides of a human torso, while Eve is positioned at various points (front, back, side). The random schedule is generated by a shared seed exchanged via an initial public‑key handshake; power levels are limited to four discrete values compatible with the hardware. Results show a packet delivery ratio above 95 % for the legitimate link, while Eve’s symbol reconstruction error remains between 48 % and 52 % across all tested positions. Even when Eve is located directly behind the body—a worst‑case geometry for line‑of‑sight channels—the system maintains a secure throughput of roughly 1 kbps, confirming the robustness of the approach for medical monitoring and other health‑care IoT applications.

The paper also discusses practical considerations. Synchronization of the random seed can be achieved during network boot‑strapping using lightweight key‑exchange protocols. Power‑level randomization requires only a small set of discrete amplitudes, which simplifies hardware implementation and does not significantly increase energy consumption. The random transmission schedule can be integrated with the existing CSMA/CA MAC of IEEE 802.15.4, preserving compatibility with legacy devices. Finally, the authors propose adaptive tuning of the transmission probability and power distribution based on real‑time channel statistics and estimated eavesdropper location to maximize the secure rate.

In summary, this work demonstrates that jointly randomizing transmission timing and power in a half‑duplex two‑way Gaussian channel yields perfect information‑theoretic secrecy with modest system overhead. The approach is analytically sound, experimentally verified, and well‑suited for low‑power, spatially constrained environments such as WBANs, offering a practical pathway to secure wireless body sensor networks without relying on higher‑layer cryptography.