Compare Papers

Paper 1

Proactive and privacy-Preserving defense for DNS over HTTPS via federated AI attestation (PAFA-DoH).

Basharat Ali, Guihai Chen

Year: 2026
Journal: Neural networks : the official journal of the International Neural Network Society
DOI: 10.1016/j.neunet.2025.108343
arXiv: -

DNS over HTTPS(DoH) improves privacy but still admits tunneling, resolver manipulation, and side-channel leakage. We present PAFA-DoH, a practical defense that unifies Federated AI, Quantum-resilient cryptography, and neuromorphic anomaly detection in a single, privacy-preserving framework. The system learns from encrypted traffic via adversarial federated learning with homomorphic inference, so models improve collaboratively without exposing raw queries. To surface hard-to-see behaviors(e.g., periodic C2 beacons), we extract topological signature from flows using persistence-diagram-based topological data analysis. Real-time detection is executed on Spiking Neural Network (SNN), running on neuromorphic hardware, delivering high throughput at low energy cost. Trust is enforced with an AI-assisted zero-knowledge attestation scheme that integrates ZK-SNARKs/STARKs to continuously verify client and resolver integrity without revealing metadata. We evaluate PAFA-DoH on a custom testbed that includes quantum-adversarial traffic, compromised-client simulations, and emulated rogue resolvers. Our Model achieves  ≥ 99.20 % malicious-activity detection accuracy,  < 100ms ZK verification, and  < 150 ms added handshake latency, while federated models converge within 10 training epochs. By combining privacy-preserving learning, verifiable trust, and event-driven analytics PAFA-DoH offers deployable path to zero-trust, post quantum-hardened protection for encrypted DNS.

Open paper

Paper 2

Toward Uncertainty-Aware and Generalizable Neural Decoding for Quantum LDPC Codes

Xiangjun Mi, Frank Mueller

Year: 2025
Journal: arXiv preprint
DOI: arXiv:2510.06257
arXiv: 2510.06257

Quantum error correction (QEC) is essential for scalable quantum computing, yet decoding errors via conventional algorithms result in limited accuracy (i.e., suppression of logical errors) and high overheads, both of which can be alleviated by inference-based decoders. To date, such machine-learning (ML) decoders lack two key properties crucial for practical fault tolerance: reliable uncertainty quantification and robust generalization to previously unseen codes. To address this gap, we propose \textbf{QuBA}, a Bayesian graph neural decoder that integrates attention to both dot-product and multi-head, enabling expressive error-pattern recognition alongside calibrated uncertainty estimates. Building on QuBA, we further develop \textbf{SAGU }\textbf{(Sequential Aggregate Generalization under Uncertainty)}, a multi-code training framework with enhanced cross-domain robustness enabling decoding beyond the training set. Experiments on bivariate bicycle (BB) codes and their coprime variants demonstrate that (i) both QuBA and SAGU consistently outperform the classical baseline belief propagation (BP), achieving a reduction of on average \emph{one order of magnitude} in logical error rate (LER), and up to \emph{two orders of magnitude} under confident-decision bounds on the coprime BB code $[[154, 6, 16]]$; (ii) QuBA also surpasses state-of-the-art neural decoders, providing an advantage of roughly \emph{one order of magnitude} (e.g., for the larger BB code $[[756, 16, \leq34]]$) even when considering conservative (safe) decision bounds; (iii) SAGU achieves decoding performance comparable to or even outperforming QuBA's domain-specific training approach.

Open paper