Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Feedback Lunch: Deep Feedback Codes for Wiretap Channels

Yingyao Zhou, Natasha Devroye, Onur Günlü

TL;DR

This work tackles secure communication over reversely-degraded Gaussian wiretap channels with channel-output feedback, where secrecy would be zero without feedback. It introduces seeded modular codes that pair universal hash-based security with a learned reliability layer (WTC-Lightcode), leveraging feedback to establish shared secret randomness and achieve positive secrecy rates. A leakage-aware training framework is developed, combining cross-entropy reliability optimization with neural estimators of information leakage and a trade-off loss that constrains leakage at a target level. The results demonstrate a practical feedback lunch effect, quantify the security–reliability trade-off across SNR regimes, and motivate secure ISAC-oriented code designs for scenarios with intrinsic feedback, while outlining future work toward larger blocklengths and hybrid code constructions.

Abstract

We consider reversely-degraded wiretap channels, for which the secrecy capacity is zero if there is no channel feedback. This work focuses on a seeded modular code design for the Gaussian wiretap channel with channel output feedback, combining universal hash functions for security and learned feedback-based codes for reliability to achieve positive secrecy rates. We study the trade-off between communication reliability and information leakage, illustrating that feedback enables agreeing on a secret key shared between legitimate parties, overcoming the security advantage of the wiretapper. Our findings also motivate code designs for sensing-assisted secure communication, to be used in next-generation integrated sensing and communication methods.

Feedback Lunch: Deep Feedback Codes for Wiretap Channels

TL;DR

This work tackles secure communication over reversely-degraded Gaussian wiretap channels with channel-output feedback, where secrecy would be zero without feedback. It introduces seeded modular codes that pair universal hash-based security with a learned reliability layer (WTC-Lightcode), leveraging feedback to establish shared secret randomness and achieve positive secrecy rates. A leakage-aware training framework is developed, combining cross-entropy reliability optimization with neural estimators of information leakage and a trade-off loss that constrains leakage at a target level. The results demonstrate a practical feedback lunch effect, quantify the security–reliability trade-off across SNR regimes, and motivate secure ISAC-oriented code designs for scenarios with intrinsic feedback, while outlining future work toward larger blocklengths and hybrid code constructions.

Abstract

We consider reversely-degraded wiretap channels, for which the secrecy capacity is zero if there is no channel feedback. This work focuses on a seeded modular code design for the Gaussian wiretap channel with channel output feedback, combining universal hash functions for security and learned feedback-based codes for reliability to achieve positive secrecy rates. We study the trade-off between communication reliability and information leakage, illustrating that feedback enables agreeing on a secret key shared between legitimate parties, overcoming the security advantage of the wiretapper. Our findings also motivate code designs for sensing-assisted secure communication, to be used in next-generation integrated sensing and communication methods.
Paper Structure (14 sections, 1 theorem, 13 equations, 5 figures, 2 tables)

This paper contains 14 sections, 1 theorem, 13 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. System Model
  3. Prior Work
  4. Proposed Coding Schemes
  5. Security Layer $(\varphi_s, \psi_s)$
  6. Reliability Layer $(e_r, d_r)$
  7. Information Leakage
  8. Trade-off Loss
  9. Reversely-Degraded Feedback Wiretap Code Simulation Results
  10. Secrecy Rate Effects
  11. Comparisons with Existing Feedback Codes
  12. Noisy Feedback Effects
  13. Security-Reliability Trade-off Loss Function Effects
  14. Conclusion

Key Result

Corollary 1

For a reversely-degraded wiretap channel with $\mathbb{P}(Y, Z|X) = \mathbb{P}(Z|X)\mathbb{P}(Y|Z)$, the secrecy capacity with noiseless feedback is

Figures (5)

  • Figure 1: Design of modular deep-learned feedback wiretap codes. The security and reliability layers are given by $(\varphi_s, \psi_s)$ and $(e_r, d_r)$, respectively.
  • Figure 2: Detailed structure of the reliability layer.
  • Figure 3: BLER (left) and estimated $\hat{I}_\text{bob}$, $\hat{L}_\text{eve}$ (right) versus blocklength under $\text{SNR} = S_{Y,f} = S_{Z,f}$ with noiseless feedback.
  • Figure 4: BLER versus forward SNR at Bob (left) and $\hat{L}_\text{eve}$ versus forward SNR at Eve (right). Each data point is obtained under the condition $\text{S}_{Y,f}=\text{S}_{Z,f}$.
  • Figure 5: Trade-off between the BLER and information leakage.

Theorems & Definitions (4)

  • Corollary 1: ahlswede2006transmission
  • Definition 1: carter1977universal
  • Definition 2: Security-advantage Gain
  • Remark 1