Encrypted communication reduces eavesdropper data leakage with adaptive methods
Leakage-Aware Encrypted Semantic Communication with Key-Conditioned Signal Perturbation
Cryptography and Security
Summary
Sending messages securely over noisy channels is tricky because eavesdroppers might intercept some information. The authors improve a system that mixes secret keys into the message to confuse eavesdroppers while still letting the intended receiver understand it. They introduce two techniques: one adds secret-key-based noise earlier in the signal, and another trains the system to minimize the amount of leaked information more effectively. Tests show these methods greatly reduce what eavesdroppers can learn without much extra cost to the real receiver.
What this means in practice
- •For secure communication engineers: Design communication systems that adaptively hide information from interceptors in wireless environments by injecting secret-key-based noise.
- •For network security teams: Deploy encryption methods that reduce data leakage to unauthorized listeners over fading wireless channels in real-time messaging systems.
Authors
Omar Kaaki, Maria Slim, Mariette Awad, Hadi Sarieddeen
Abstract
Encrypted 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. We propose two changes to the encrypted semantic communication system (ESCS): a key-conditioned signal-domain perturbation layer, called learned adversarial perturbation injection (LAPI), and an information noise-contrastive estimation (InfoNCE)-guided adaptive training rule that {adds an adaptive leakage penalty alongside the fixed adversarial objective} using an InfoNCE-based proxy for mutual information leakage. Experiments on the Europarl corpus over additive white Gaussian noise (AWGN) and Rayleigh fading channels show that InfoNCE-guided training is the primary source of eavesdropper suppression, reducing the eavesdropper's bilingual evaluation understudy (BLEU) score to approximately 0.07 at high signal-to-noise ratio (SNR). Adding LAPI to the InfoNCE-guided objective further lowers the InfoNCE-estimated leakage by 5-9% with low measured inference overhead, while its impact on legitimate-receiver BLEU depends on channel conditions. The two mechanisms operate on different parts of the pipeline and can be jointly applied.