Neural Network Verification for Deep Joint Source-Channel Coding

2026-10-08

AI Summary

This study addresses the vulnerability of deep joint source-channel coding (DeepJSCC) to adversarial perturbations, where reconstruction quality degrades sharply without guaranteed error bounds. We propose the first verification framework for DeepJSCC decoders. By extending linear relaxation, structural encoding, and GloRo training, our approach achieves global robustness certification supporting components such as PReLU for the first time. Furthermore, we introduce constrained upsampling convolutions, structured perturbation modeling for Rayleigh fading, and Lipschitz regularization to rigorously bound worst-case reconstruction errors under wireless noise. Experimental results demonstrate that the median certified bound is reduced by 41%, while verified safe cases increase approximately tenfold. The empirically observed maximum error reaches only 0.082, substantially below the certified upper bound of 0.128, confirming the tightness and practical effectiveness of the proposed framework.

Abstract

Deep joint source-channel coding (DeepJSCC) transmits data end-to-end over wireless channels using a neural encoder-decoder, but reconstruction quality can degrade sharply under adversarial perturbations and channel disturbances; no method formally bounds this degradation for DeepJSCC. We present the first bound-propagation framework for verifying DeepJSCC's decoder, bounding worst-case reconstruction error over a given wireless channel's noise region. Current deep neural network (DNN) verifiers do not support three DeepJSCC decoder components: parametric rectified linear activations (PReLU), transposed convolutions, and Rayleigh fading. We extend state-of-the-art techniques for optimization of linear relaxation in DNN verification for PReLU, replace the transposed convolution with its restricted upsample-then-convolution form, and formulate Rayleigh fading as a structural perturbation prepended directly into the decoder, thereby reducing the dimensionality of the verification problem. We also instantiate Lipschitz-regularized global robustness training, denoted GloRo, improving global robustness and enabling tight certification of DeepJSCC models for the first time. On DeepJSCC model for image transmission, this global robustness training procedure combined with structural encoding lowers the median certified bound by up to 41% and certifies about ten times more safe cases (192 against 19) than GloRo with interval encoding at a 10-degree error in channel estimation. Over-the-air validation with an orthogonal frequency-division multiplexing (OFDM) implementation on software-defined radio devices confirm the certificate holds on real hardware, with a worst observed error on radio link at 0.082 against a certified bound of 0.128.

Problem

• Deep Joint Source-Channel Coding
• Neural Network Verification
• Reconstruction Error Bound
• Wireless Channel Noise
• Adversarial Perturbations

Innovation

• DeepJSCC
• Neural Network Verification
• Bound Propagation
• Global Robustness Training
• Rayleigh Fading