Introducing Feisty Duck News: Cryptography and Security, Live

Short News

Artificial Intelligence

• Advertisers are trying to influence AI bots with secret ads
  A German researcher discovered AI-specific advertisements embedded in feeds targeted at AI crawlers, appearing as extensive FAQs visible only to AI agents, not human users. The article also discusses tensions between OpenAI and Hugging Face, as well as the rise of competitive Chinese open-weight AI models, such as DeepSeek and Alibaba's offerings. The hidden ads raise concerns about data integrity, publisher trust, and potential model poisoning through biased training data.
• VMs won't contain cyber-capable agents
  Trail of Bits reports that consumer-grade virtual machines are insufficient to contain modern AI agents. Testing showed GPT-5.6-Cyber escaped sandboxed QEMU/KVM VMs three times, autonomously discovering three zero-days and building working exploits. The report recommends using hardened virtualization with a minimal attack surface, such as Firecracker, to provide stronger containment of advanced agents.
• How We Secure Figma's Internal Systems With Agents
  Figma describes its security operations approach using AI agents built on a RAG stack to automate alert triage, forensic investigations, and incident response. The system routes security alerts through specialized LLM-powered agents with scoped access to tools like Okta, Wiz, Panther, and Snowflake, reducing alert resolution time by 71% while maintaining security controls through agent specialization and limited permissions.

Post-Quantum Cryptography

• RFC 10024: Negotiating Post-Quantum TLS handshakes
  RFC 10024 introduces hybrid key exchange mechanisms to integrate post-quantum cryptography into TLS 1.3, using ML-KEM-768. The standard preserves existing TLS mechanics while enabling incremental deployment through HelloRetryRequest negotiation, addressing both passive decrypt-later attacks and active quantum threats. Implementation support already exists in OpenSSL and other common TLS stacks, with a roadmap to make hybrid the default by 2029.
• The ePrint:2026/1591 Quantum Algorithm Does Not Solve DCP
  A critique of Simon's claimed quantum algorithm for the Dihedral Coset Problem (DCP), proving that algorithms discarding too much information about Fourier-label coordinates cannot solve DCP in polynomial time. The authors establish a general no-go result and suggest that successful DCP solutions require better utilization of Fourier label information.
• Quasipolynomial Cryptanalysis of the McEliece Cryptosystem (or: PIR Meets McEliece)
  A paper presenting a quasipolynomial-time distinguisher and attack against the McEliece cryptosystem using binary Goppa codes. The attack runs in time n^O(log n) and can distinguish McEliece public keys from random matrices with high probability, applying to all Classic McEliece parameter sets evaluated in the NIST standardization process. The approach is extended heuristically to recover messages from noisy codewords, though results are not yet practically feasible.
• Europe's Post-Quantum Readiness 2026: An Empirical Assessment of the EU-27
  A Zenodo report assessing EU-27 countries' preparedness for the migration to post-quantum cryptography using a five-tier framework. Findings show Lithuania alone meets the highest tier (strong government-led PQC implementation), with 44.4% of countries in the top two tiers and 81.5% showing some structured PQC preparation. The report identifies a significant gap between EU policy intent and actual verifiable implementation across member states.
• Post-Quantum Cryptography in Long-Term Support JDK Releases
  Oracle is extending post-quantum cryptography support across Java's long-term support (LTS) releases, including JDK 25, 21, 17, 11, and 8. Starting with ML-KEM and ML-DSA introduced in JDK 24, and hybrid PQC key exchange for TLS 1.3 in JDK 27 (planned September 2026), Oracle plans to backport these capabilities to older LTS releases through 2027, enabling gradual enterprise migration with minimal disruption.
• Post-quantum JWT signing: ML-DSA is here
  The article explains post-quantum signing for JSON Web Tokens using ML-DSA (Dilithium), standardized in RFC 9964. It covers three ML-DSA variants (ML-DSA-44, ML-DSA-65, ML-DSA-87), the new AKP JWK type for ML-DSA keys, and practical considerations, including larger signature sizes. The article recommends adopting ML-DSA for long-lived tokens, high-stakes security, or when customers require post-quantum support, while noting that classical algorithms remain adequate for short-lived tokens.
• NSA and IETF, part 9: An update
  The post critiques IETF standardization procedures for a hybrid ECC+ML-KEM specification in TLS, arguing that working group chairs declared rough consensus to advance the specification despite lacking majority support among participants and leaving key objections unaddressed. The author documents concerns about procedural integrity and governance transparency in IETF decision-making.

Cryptography

• Signal: Introducing Automatic Key Verification
  Signal introduces Automatic Key Verification, a feature that provides streamlined end-to-end encryption verification through a layered system of user checks, Signal connections, and third-party auditors. It uses key transparency and ledger-based processes to ensure public key consistency without requiring in-person meetings or secondary communication channels.
• RFC 10015: Deprecating Obsolete Key Exchange Methods in TLS 1.2 and DTLS 1.2
  RFC 10015 deprecates obsolete key exchange methods in TLS 1.2 and DTLS 1.2, including non-ephemeral Diffie-Hellman and ECDH, FFDHE, and RSA-based cipher suites. The RFC recommends implementations move toward stronger, ephemeral, and non-RSA-based key exchanges while noting that TLS 1.3 is unaffected.
• John Gordon: The Alice and Bob After-Dinner Speech
  A humorous 1984 overview of coding theory presented as a 'biography' of Alice and Bob, covering source coding (data compression), channel coding (error correction), and secrecy coding (cryptography), with practical examples and trade-offs in reliable and secure communication.
• echtool-web: Probe ECH deployments with customizable configurations
  echtool-web is a web interface for testing and probing Encrypted Client Hello (ECH) deployments. It offers customizable configuration options, validation of RetryConfigurations, and CLI-like commands for sending GREASE ECH requests, establishing connections with ECH configs, and decoding ECH configurations. The web interface supports IPv4 only; IPv6 is available through the standalone echtool CLI.
• IC researchers win Internet Defense Prize for new encryption method - EPFL
  EPFL researchers developed E2EE-CD, a new end-to-end encryption method for collaborative document editing that extends the Signal protocol to work as an encrypted broadcast channel. The system enables secure multi-user collaboration without exposing content to service providers, achieving approximately 120 ms latency while maintaining scalability and supporting role-based access control.

Privacy and Society

• What are the best private browsers in 2026?
  A review and ranking of privacy-focused web browsers for 2026, comparing desktop browsers under default settings. The article evaluates privacy mechanisms like state partitioning to prevent cross-site tracking. It discusses how features such as Alt-Svc and blob URLs affect user privacy, including specific version information for browsers such as Brave, Firefox, Ungoogled Chromium, Tor, and LibreWolf.
• ShaderGhost — the tracking ID you can't delete
  ShaderGhost demonstrates a covert tracking mechanism that exploits GPU shader caches to create persistent identifiers that persist even after standard privacy actions such as clearing cookies or restarting the browser. The tracking ID is stored locally on the device via WebGL shader compilation timing side channels and persists across sites, evading typical browser privacy controls. The article advocates that browsers partition or clear GPU caches, similar to how they handle cookies.
• Privacy Thing
  Privacy Thing is a browser extension that reduces online fingerprinting by shielding 53 browser properties, methods, and constructors across 13 categories, including geolocation, Canvas, WebGL, WebRTC, and others. Available as a preview for Chrome, Firefox, and Edge.
• Bad Connection: Uncovering Global Telecom Exploitation by Covert Surveillance Actors
  The Citizen Lab investigates covert surveillance actors exploiting weaknesses in global 4G/5G telecom networks and international roaming infrastructure to track targets. The research identifies sophisticated threat actors who manipulate SS7 and Diameter signaling protocols, spoof operator identities across multiple countries, and use IPX providers to conceal their activities, revealing systemic vulnerabilities in telecom infrastructure.
• Sensitive Info Goes Into 'No Reply' Emails Constantly. This Guy Sees It All
  Security researcher Cory Solovewicz receives hundreds of thousands of emails daily sent to misconfigured 'no reply' addresses, exposing sensitive information through poorly managed email systems. The article highlights how leaked contact addresses, bulk email practices, and inadequate data disposal allow confidential details to end up in unmonitored inboxes, demonstrating systemic failures in email security and data handling.

Public Key Infrastructure

• The SPIFFE Standard Roadmap
  The post outlines the SIG-Spec working group's plans to evolve the SPIFFE standard over the next 12 months, with an emphasis on community engagement and industry-prioritized features. It details weekly open meetings for discussion and proposals, inviting researchers, practitioners, and enthusiasts to participate through the SPIFFE community.
• EU Age Verification Project Mandates Hardware-Bound Attestation
  An EU-funded open-source age-verification project requires hardware-bound attestation using trusted hardware (Android TEE, Apple Secure Enclave) to prevent credential copying. The approach raises concerns about device dependency and ecosystem lock-in, while credentials are restricted to Commission-approved applications. A security review is forthcoming to evaluate alternative approaches.
• CertSight: Realtime certificate monitoring via eBPF
  CertSight is an eBPF-based real-time certificate observability tool for Linux that monitors TLS certificate activity across the infrastructure without accessing private keys or requiring application changes. It supports cloud-native, bare metal, and Kubernetes environments, detects certificates via system file access and network events, includes Java support, and provides detailed observability, including TLS versions and cipher suites.

Security

• ENISA Secure by Design and Default Playbook
  ENISA has published a practical guide for SMEs to implement secure-by-design and secure-by-default principles throughout a product's lifecycle. The playbook includes 22 actionable playbooks designed to integrate these security concepts into existing engineering, product, and release processes, with materials available on GitHub.
• CSS: the bomb inside your inbox
  PortSwigger Research demonstrates multiple CSS and HTML sanitization bypass techniques in webmail clients (Gmail, Outlook, ProtonMail, Yahoo Mail, AOL Mail, Fastmail, Slack). Attackers can exploit divergences between sanitizers and browser rendering to perform UI spoofing, exfiltrate authentication tokens, track email views, and capture keystrokes, potentially leading to account takeover.
• Everything I Learned Shipping Device Bound Session Credentials
  A detailed account of implementing Device Bound Session Credentials (DBSC) at Report URI, covering the practical challenges and lessons learned. DBSC binds session cookies to hardware-backed device keys (TPM/secure enclave) with a two-phase refresh workflow. Key lessons include: proper state management to avoid race conditions, mandatory cookie rotation on refresh, avoiding registration response challenges that could cause browser errors, and handling browser-specific quirks such as transient cookie mismatches and unintended logouts.
• The Devil Is In The Defaults: What To Do About XSS
  This talk explores XSS defense mechanisms and browser-level protections, including Trusted Types policies, setHTML as a safer alternative to innerHTML, and policy-free browser sanitization to prevent XSS attacks. The speaker discusses the current implementation status across browsers and envisions a future in which XSS is mitigated by default through standardized browser mechanisms, analogous to the adoption of HTTPS.
• BGP Role model: tracking the adoption of RFC 9234
  Cloudflare analyzes the deployment of RFC 9234, which introduces BGP Role and the Only to Customer (OTC) path attribute to prevent route leaks. The study finds partial adoption with some Tier-1 networks stripping OTC attributes, hindering propagation. The article provides guidance on local deployment of BGP Roles and enabling strict mode for stronger route-leak prevention.
• Beyond origin validation: Four classes of routing attack nobody is validating
  The article presents a taxonomy of four macro-categories of BGP routing attacks that lack robust validation mechanisms: route manipulation, routing consistency, policy violation, and session-based attacks. While origin validation via RPKI/ROV addresses some threats, the article identifies significant gaps in protection for ASM (only partially addressed by undeployed BGPsec), RLK (showing promise with ASPA objects), and VOL/DEG/POL/ATR attacks (lacking cryptographic protections and relying on operational hardening). The analysis suggests that as origin validation improves, attackers may shift toward these unvalidated attack classes.
• The State of Passkeys: Studying the Adoption and Security of Passkeys on the Web
  A large-scale study analyzing passkey deployment across 872 websites using the PASSKEYS-RADAR dataset. Researchers discovered inconsistent passkey management practices and the use of deprecated cryptographic algorithms, and demonstrated 15 attack types (10 novel) using the PASSKEYS-ATTACKER tool. Testing revealed 53 of 103 sites vulnerable to high-severity attacks, including account takeover, phishing, and session fixation.

Classifieds

• Security Engineer (L5) - Workforce Security | Netflix. We are seeking a talented L5 Security Engineer specializing in Endpoint Security to join our team. NETFLIX
• Security Engineer, Insider Threat Detection & Response | OpenAI. This role will focus on all aspects of Detection & Response but with a strong emphasis on detecting insider threats and influencing controls to safeguard OpenAI's most sensitive assets. OPENAI
• Lead Product Manager, Security | Wikimedia Foundation. We are hiring a Lead Product Manager to bring their experience dealing with modern threats to bear on Wikipedia’s security strategy in the AI era, and to drive a strong security culture throughout the Wikimedia Foundation and movement. WIKIMEDIA
• Director of Quantum and Advanced Computing | HSBC. Set the Group-wide QAC strategy and roadmap across quantum and advanced computing aligned to HSBC priorities. HSBC

We use Claude to help us create the short news section.