FINMA tells Swiss banks to inventory crypto and get quantum-safe, after finding most have no migration roadmaps.
trending30
31 32 Can federated learning keep its accuracy once you bolt on HE and DP, or does client heterogeneity still wreck it?33 Report on an attested TLS relay flaw hitting WhatsApp and Cocos AI, proof that handshakes still cosplay as security.34 PromptGraph: Graph-Guided Prompt Sanitization for Balancing Privacy and Utility in LLM Inference arxiv.orgCan a graph over prompt spans scrub PII without torching LLM utility?35 Ammolite Security and Myntex Showcase Post-Quantum Secure Communications at NATO SHINE 2025 — by Myntex Inc youtu.beYouTube video of Myntex/Ammolite’s NATO SHINE 2025 demo of post-quantum secure comms, because quantum fear sells hardware.36 Paper tests watermark-like logit perturbations against black-box LLM theft, then measures how much utility gets shaved off.37 Can an LLM write EasyCrypt proofs, and this repo ships the MCP workflow, benchmarks, and replay logs to find out.38 Front-running meets a power-weighted lottery here, and the paper claims that makes sandwiching uneconomic.39 Devil in the Lens: Analyzing and Defending Physical Prompt Injection Against Vision-Language Models on Wearable Devices arxiv.orgPhysical prompt injection on AI glasses works across VLMs; the paper also benchmarks a mask filter and semantic detector.40 Byzantine Accountability Without Consensus: Strong Eventual Consistency for Non-Associative, Stochastic, Robust Aggregation arxiv.orgACFA: consensus-free Byzantine replication for robust aggregators like multi-Krum, by CRDTing signed contributions and equivocations.41 A post-quantum migration pitch that, sensibly, starts with inventory and CBOMs instead of swapping algorithms in a panic.42 Need the docs for PODs, GPCs, and ZApps, aka programmable cryptography over user-owned data?43 Need a machine-checked PCS spec and KZG proofs in Isabelle, or just the games and assumptions? This AFP entry has both.44 A Security magazine recap of PQC adoption, NIST standards, CNSA 2.0 timelines, and the usual harvest-now-decrypt-later anxiety.45 Report on ZWEICHAIN-Q, a PQ finality/evidence layer with real signature tests, deterministic roots, and restart metrics.46 LLM-guided evolution breaks perceptual hashes like pHash, PDQ, PhotoDNA, and NeuralHash in black-box mode, because apparently those work...47 Google's 2029 quantum crypto warning is just the old Shor panic with a date stamp and a blockchain angle.48 Can you trust Opal2 SEDs on Linux? This black-box study of 38 drives says firmware bugs and incompatibilities say no.49 Cardano's governance paper is a full spec plus research wish list for proving the thing doesn't deadlock.50 The hard part isn’t PQC, it’s finding every stray RSA/ECC dependency before the migration spreadsheet explodes.51 FIPS 140 freezes the crypto stack in amber, so you inherit old, opaque code and the bugs in its HSM plumbing.52 Survey of LLM watermarking, from biased sampling to adaptive-removal attacks, because even text wants DRM.53 A10 says ACOS 7.0.3 adds hybrid PQ TLS with OpenSSL 3.5, so your ADC can do X25519+ML-KEM without new boxes.54 qorechain/qorechain-pqc: Post-quantum cryptography for QoreChain — standardized NIST primitives (ML-DSA / ML-KEM / SHAKE-256) in six languages, proven byte-compatible against shared test vectors. github.comMulti-language wrappers for NIST PQC primitives, with shared test vectors so six bindings can agree on bytes for once.55 Anonymous comms on constant-degree expanders, via onion routing and random walks, match complete-network asymptotics.56 HANNS: Low-Storage Non-Interactive Approximate Private Nearest Neighbor Search with Sublinear Comparison Complexity eprint.iacr.orgSublinear encrypted comparisons for PNN search, by clustering and homomorphic PQ, because fully homomorphic pain needed pruning.57 Can TFHE bootstrapping get cheaper by splitting lookup tables and doing homomorphic div/rem?58 Attested TLS Was Supposed to Be the Last Trust Boundary. It Isn't. Formal Methods Show How. hackernoon.comProVerif finds CVE-2026-33697, a relay attack that breaks attested TLS binding, so the last trust boundary wasn't.59 Trace-only segmentation of cipher implementations, no labels or metadata, just repetition-scale estimates and stable cores.60 Weak-key distinguishers pop out of a split-and-cancel oracle, extending integral attacks to SIMON, SPECK, PRESENT, and GIFT.