Key-conditioned perturbation reduces eavesdropper leakage in semantic communication
Original title: Leakage-Aware Encrypted Semantic Communication with Key-Conditioned Signal Perturbation
InfoNCE-guided training suppresses eavesdropper BLEU score to 0.07 while LAPI adds 5-9% leakage reduction.
The proposed ESCS inserts a key-conditioned signal-domain perturbation layer called LAPI into the deep-learning pipeline. An InfoNCE-guided adaptive training rule adds an adaptive leakage penalty to the fixed adversarial objective. Experiments on Europarl over AWGN and Rayleigh fading channels demonstrate primary suppression from InfoNCE training and further improvement when mechanisms are combined.
- Eavesdropper BLEU score reaches ~0.07 at high SNR
- LAPI lowers InfoNCE-estimated leakage by 5-9%
- Mechanisms operate on different pipeline parts
- Joint application yields low inference overhead
From the original article
1 min read at arxiv.orgEncrypted semantic communication systems protect transmitted messages by inserting key-conditioned modules into a deep-learning semantic communication pipeline, but existing designs rely on a fixed adversarial weight and operate mainly at the semantic-feature level.
By Omar Kaaki · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.