HU.HN

Infrastructure overview

HU.HN

A quick look at the public, non-classified part of our infrastructure. For questions, or for the non-public, more detailed overview, get in touch. That version, available to customers under a binding NDA, covers our active-defense measures: deception infrastructure, disguised honeypots and reactive countermeasures against attackers.

Public edition Not classified October 2026
3authoritative nameservers, operated worldwide
DNSSECzones signed by default
3mail nodes in a redundant cluster
Rustmemory-safe mail stack

01 / DNS

Authoritative DNS, worldwide

The base of the hu.hn network is our globally distributed, highly available and non-recursive nameserver infrastructure. Every name resolution ends on one of these three hosts:

pegasus.hu.hn
cassiopeia.hu.hn
andromeda.hu.hn
  • Signed zones by default. DNSSEC protects resolution against spoofing, cache poisoning and related attack patterns.
  • Authoritative only. The nameservers answer for our zones and nothing else. No recursion means a smaller attack surface and no use as an open resolver in amplification attacks.
  • Modern cryptography. Post-quantum protections are part of our cryptographic baseline, so today's traffic stays protected tomorrow.

Planned: anycast routing

Today, a resolver chooses among the three nameservers itself. We are working on GRE-based anycast, which will route each query to the nearest data center with a running instance and fail over immediately when one goes down. Talks with several carriers are already under way.

02 / Mail

Highly redundant, modern mail

Mail is our next primary service. It runs on Stalwart, a modern mail and collaboration server written in memory-safe Rust, in a highly redundant cluster with direct delivery to all major mail providers. The Stalwart codebase has been independently security-audited.

iris.hu.hn
hermes.hu.hn
apollo.hu.hn

Three servers, each handling mail. Two run the complete service stack. The third acts as the witness: it does not carry the full stack, its job is to cast the deciding vote. Cluster decisions follow majority voting (quorum): a failover only happens when more than half of the voters agree. With only two nodes, a network split leaves each side unable to tell whether its peer has failed or is merely unreachable, so both could carry on independently (split brain). With three voters, at most one side of any split can hold a majority, so there is always exactly one authoritative side. This is the principle behind consensus algorithms such as Paxos and Raft. Any node can serve any protocol, and the delivery queue is shared, so another node takes over if one fails.

Protocols & clients

JMAP, IMAP4rev2/rev1, POP3, SMTP and ManageSieve for server-side filter rules. Works with Apple Mail, Thunderbird, Outlook and mobile clients, with automatic client configuration. CalDAV, CardDAV and WebDAV cover calendars, contacts and files.

Sender authentication

SPF, DKIM, DMARC and ARC. DKIM keys are generated, published and rotated automatically; incoming DMARC and TLS reports are ingested and visualised.

Transport security

DANE, MTA-STS and TLS reporting protect server-to-server delivery against downgrade attacks. TLS certificates are issued and renewed automatically via ACME.

Encryption at rest

Mailboxes can be encrypted with the user's own S/MIME certificate or OpenPGP key, so even an operator with disk access cannot read them.

Spam & phishing defense

Built-in statistical classifier, DNS blocklists, sender reputation, greylisting, spam traps and Pyzor digests, plus protection against homograph and spoofing attacks. Optional LLM-based analysis.

Account security

Two-factor authentication (TOTP), app passwords, OAuth 2.0 / OpenID Connect, access control lists, rate limiting and automatic blocking of abusive IP addresses.

03 / Platform security

Security from the silicon up

These are our baseline security requirements for every system we operate. They start at the hardware, long before any software, customer or workload comes into play.

Dedicated hardware only

No shared hosts and no shared tenancy underneath us. We run on dedicated servers with current AMD Zen 4 and Zen 5 processors, which bring the hardware security features everything else builds on.

Encrypted at rest and in use

Quantum-resistant storage encryption is combined with hardware memory encryption (AMD SME/TSME, plus SEV-SNP for virtual machines), so data stays protected on disk and in RAM.

Nothing unlocks on unverified firmware

Before a single encryption key is even considered for release, the platform's firmware state is verified against a reference that has been personally vetted twice. If anything differs, nothing is unlocked.

Physical access does not help

Evil-maid-style attacks are part of our threat model. A server that is grabbed and powered on elsewhere is just a box of unusable data. An attempt to infiltrate one is detected, and we act on it.

State-level adversaries are in scope

Our threat model explicitly includes well-resourced actors, including intelligence services and other state actors. Access to systems or data does not happen around our review, and attempts to bypass it are noticed.

Dispersed storage, under evaluation

We are evaluating AONT-RS, an all-or-nothing transform combined with Reed-Solomon coding. Data is split into fragments that are individually meaningless, so even a compromised host is, as far as possible, a piece of junk.

Real isolation, no shortcuts

Where we virtualize, we do it properly: an extremely hardened stack with hardware-assisted isolation. A container is not isolation, just a convenient way to organize processes, so we do not use containers as a security boundary, and customer workloads never share a kernel.

The design goal

Grab a server and run, and all you get is unusable data. Try to infiltrate one, and we will know and act on it.

04 / Compliance

Not a checkbox. A moving target we follow.

We are not joking around here. The rules for hosting, DNS and mail change constantly, so we track legal and regulatory changes worldwide on an ongoing basis instead of waking up when an audit date gets close. More than 45 laws and standards are on our radar, across the EU, Germany, the UK, Switzerland, North America and beyond. NIS2 is the baseline for us, not the finish line: we build well beyond what it defines as the minimum.

Continuous monitoring

Changes in EU, German and international law are followed as they happen, not once a year before an audit.

From law to control

Every relevant requirement is translated into a concrete technical control and checked against our infrastructure.

Beyond the baseline

Minimum requirements set the floor. Our internal standards sit above them, in security as well as in privacy.

European Union

  • GDPRRegulation (EU) 2016/679
  • ePrivacy Directive2002/58/EC
  • NIS2Directive (EU) 2022/2555, incl. Art. 28
  • DORARegulation (EU) 2022/2554
  • Cyber Resilience ActRegulation (EU) 2024/2847
  • EU AI ActRegulation (EU) 2024/1689
  • Data ActRegulation (EU) 2023/2854
  • Digital Services ActRegulation (EU) 2022/2065
  • e-EvidenceRegulation (EU) 2023/1543
  • eIDAS 2.0Regulation (EU) 2024/1183
  • EUCCImplementing Reg. (EU) 2024/482
  • EU-US Data Privacy FrameworkDecision (EU) 2023/1795

Germany

  • BDSGFederal Data Protection Act
  • TDDDGformerly TTDSG
  • TKGTelecommunications Act
  • NIS2UmsuCG / BSIGGerman NIS2 implementation
  • KRITIS-Dachgesetzphysical resilience (CER)
  • BSI IT-GrundschutzCompendium
  • BSI C5cloud compliance catalogue
  • BSI TR-03108secure e-mail transport
  • BSI TR-02102cryptographic mechanisms
  • TISAXautomotive assessment scheme

United States

  • CCPA / CPRACalifornia
  • BIPAIllinois, 740 ILCS 14
  • ECPAelectronic communications
  • HIPAAhealth data
  • COPPAchildren's data
  • GLBAfinancial data
  • FERPAeducation records
  • SOXfinancial reporting
  • CLOUD Actcross-border data access

UK, Switzerland & rest of world

  • UK GDPRand Data Protection Act 2018
  • Swiss revFADPrevised data protection act
  • LGPDBrazil
  • APPIJapan
  • PIPLChina
  • PIPEDACanada
  • PDPASingapore

Standards & frameworks we measure our controls against

  • ISO/IEC 27001:2022information security
  • ISO/IEC 27701privacy information management
  • ISO/IEC 42001AI management systems
  • SOC 2trust services criteria
  • PCI DSS 4.0.1payment card security
  • NIST SP 800-53 Rev. 5security and privacy controls
  • NIST CSF 2.0cybersecurity framework
  • CIS Controls v8hardening baseline
  • NIST FIPS 203 / 204 / 205post-quantum cryptography

Laws and regulations listed here are tracked and mapped to our controls. Standards listed here serve as reference frameworks for our controls. This overview describes our approach and is not a certification statement.

05 / Beyond this overview

Beyond what is listed here

Active defense, under NDA

The measures above are the ones we can describe publicly. Our operational security goes further, with deception infrastructure, disguised honeypots and reactive countermeasures against attackers. These details are available on request to our customers after signing a strictly binding non-disclosure agreement.

06 / Standards & sources

Standards, specifications & law

The RFCs, standards and legal texts behind this overview, for anyone who wants to check our work.

DNS & DNSSEC
  • RFC 1034Domain Names: Concepts and Facilities
  • RFC 1035Domain Names: Implementation and Specification
  • RFC 8499DNS Terminology
  • RFC 4033DNS Security Introduction and Requirements
  • RFC 4034Resource Records for the DNS Security Extensions
  • RFC 4035Protocol Modifications for the DNS Security Extensions
  • RFC 5155DNSSEC Hashed Authenticated Denial of Existence (NSEC3)
  • RFC 6781DNSSEC Operational Practices, Version 2
  • RFC 9364DNS Security Extensions (DNSSEC) (BCP 237)
  • RFC 9904DNSSEC Cryptographic Algorithm Recommendation Update Process (obsoletes RFC 8624)
  • RFC 1996A Mechanism for Prompt Notification of Zone Changes (DNS NOTIFY)
  • RFC 5936DNS Zone Transfer Protocol (AXFR)
  • RFC 8945Secret Key Transaction Authentication for DNS (TSIG)
  • RFC 7873Domain Name System (DNS) Cookies
  • RFC 4786Operation of Anycast Services (BCP 126)
  • RFC 2784Generic Routing Encapsulation (GRE)
  • RFC 9276Guidance for NSEC3 Parameter Settings (0 iterations, empty salt)
  • SP 800-81r3Secure Domain Name System (DNS) Deployment Guide (NIST, March 2026)
Mail protocols
  • RFC 5321Simple Mail Transfer Protocol
  • RFC 5322Internet Message Format
  • RFC 9051Internet Message Access Protocol (IMAP), Version 4rev2
  • RFC 1939Post Office Protocol, Version 3
  • RFC 8620The JSON Meta Application Protocol (JMAP)
  • RFC 8621The JSON Meta Application Protocol (JMAP) for Mail
  • RFC 5228Sieve: An Email Filtering Language
  • RFC 5804A Protocol for Remotely Managing Sieve Scripts (ManageSieve)
  • RFC 4791Calendaring Extensions to WebDAV (CalDAV)
  • RFC 6352CardDAV: vCard Extensions to WebDAV
  • RFC 4918HTTP Extensions for Web Distributed Authoring and Versioning (WebDAV)
Mail authentication & transport security
  • RFC 7208Sender Policy Framework (SPF)
  • RFC 6376DomainKeys Identified Mail (DKIM) Signatures
  • RFC 9989Domain-based Message Authentication, Reporting, and Conformance (DMARC) (obsoletes RFC 7489)
  • RFC 9990DMARC Aggregate Reporting
  • RFC 9991DMARC Failure Reporting
  • RFC 8617The Authenticated Received Chain (ARC) Protocol
  • RFC 3207SMTP Service Extension for Secure SMTP over TLS (STARTTLS)
  • RFC 8314Cleartext Considered Obsolete: Use of TLS for Email Submission and Access
  • RFC 6698DNS-Based Authentication of Named Entities (DANE) TLSA
  • RFC 7672SMTP Security via Opportunistic DANE TLS
  • RFC 8461SMTP MTA Strict Transport Security (MTA-STS)
  • RFC 8460SMTP TLS Reporting
  • RFC 8555Automatic Certificate Management Environment (ACME)
Encryption, identity & cryptography
  • RFC 9580OpenPGP
  • RFC 8551S/MIME Version 4.0 Message Specification
  • RFC 6238TOTP: Time-Based One-Time Password Algorithm
  • RFC 6749The OAuth 2.0 Authorization Framework
  • OIDC Core 1.0OpenID Connect Core 1.0 (OpenID Foundation)
  • FIPS 203Module-Lattice-Based Key-Encapsulation Mechanism (ML-KEM) (NIST)
  • FIPS 204Module-Lattice-Based Digital Signature Standard (ML-DSA) (NIST)
  • FIPS 205Stateless Hash-Based Digital Signature Standard (SLH-DSA) (NIST)
Platform security & hardware
  • AMD SEVSecure Encrypted Virtualization: SEV, SEV-ES, SEV-SNP (AMD)
  • SP 800-193Platform Firmware Resiliency Guidelines (NIST)
  • TPM 2.0TPM 2.0 Library Specification (ISO/IEC 11889) (Trusted Computing Group)
Law, regulation & frameworks
  • GDPRRegulation (EU) 2016/679
  • ePrivacyDirective 2002/58/EC
  • NIS2Directive (EU) 2022/2555
  • DORARegulation (EU) 2022/2554
  • CRACyber Resilience Act, Regulation (EU) 2024/2847
  • AI ActRegulation (EU) 2024/1689
  • Data ActRegulation (EU) 2023/2854
  • DSADigital Services Act, Regulation (EU) 2022/2065
  • e-EvidenceRegulation (EU) 2023/1543, European Production and Preservation Orders
  • 2023/1544Directive (EU) 2023/1544, designated establishments and legal representatives
  • eIDAS 2.0Regulation (EU) 2024/1183
  • EUCCImplementing Regulation (EU) 2024/482, EU Common Criteria-based certification scheme
  • DPFImplementing Decision (EU) 2023/1795, EU-US Data Privacy Framework
  • CERDirective (EU) 2022/2557, resilience of critical entities (basis of the KRITIS-Dachgesetz)
  • NIS2UmsuCGBGBl. 2025 I Nr. 301, German NIS2 implementation act (in force since 6 Dec 2025)
  • BDSGBundesdatenschutzgesetz
  • TDDDGTelekommunikation-Digitale-Dienste-Datenschutz-Gesetz
  • TKGTelekommunikationsgesetz
  • NIST 800-53Security and Privacy Controls, Rev. 5
  • NIST CSF 2.0Cybersecurity Framework 2.0

07 / Research

The research we stand on

The papers, write-ups and advisories below shaped our design. Several of them show exactly where hardware protections end, and we included them on purpose: our threat model starts from those limits, not from the marketing sheet. Quorum voting is defined in the distributed-systems literature rather than in an RFC, so it lives here too.

Physical memory attacks
  • Cold bootJ. A. Halderman, S. D. Schoen et al.: Lest We Remember, Cold Boot Attacks on Encryption Keys (USENIX Security, 2008)
  • ScramblersS. F. Yitbarek, M. T. Aga, R. Das, T. Austin: Cold Boot Attacks are Still Hot, Security Analysis of Memory Scramblers in Modern Processors (IEEE HPCA, 2017)
  • DRAM remanence3mdeb: Conclusions from RAM data remanence tests on DDR4 and DDR5 (2025)
  • Cold boot defenseKicksecure: Cold Boot Attack Defense, summarizing the 3mdeb measurements (wiki)
  • Memory side channelsMemory Under Siege: A Comprehensive Survey of Side-Channel Attacks on Memory (arXiv:2505.04896, 2025)
  • BadRAMJ. De Meulemeester, L. Wilke, D. Oswald, T. Eisenbarth, I. Verbauwhede, J. Van Bulck: BadRAM, Practical Memory Aliasing Attacks on Trusted Execution Environments (IEEE S&P, 2025; CVE-2024-21944)
  • Battering RAMJ. De Meulemeester, D. Oswald, I. Verbauwhede, J. Van Bulck: Low-Cost Interposer Attacks on Confidential Computing via Dynamic Memory Aliasing (IEEE S&P, 2026)
  • WireTapA. Seto et al.: Breaking Server SGX via DRAM Bus Interposition (ACM CCS, 2025)
Confidential computing & CPU isolation
  • SEV-SNPD. Kaplan: SEV-SNP, Strengthening VM Isolation with Integrity Protection and More (AMD white paper, 2020)
  • Intel TMEIntel Architecture Memory Encryption Technologies (TME, TME-MK) (Intel specification)
  • SEV glitchR. Buhren et al.: One Glitch to Rule Them All, Fault Injection Attacks Against AMD SEV (ACM CCS, 2021)
  • InceptionD. Trujillo, J. Wikner, K. Razavi: Exposing New Attack Surfaces with Training in Transient Execution (USENIX Security, 2023)
  • Enter, ExitO. Oleksenko et al.: Enter, Exit, Page Fault, Leak, Testing Isolation Boundaries for Microarchitectural Leaks (IEEE S&P, 2026)
Boot integrity, TPM, DMA & hypervisor isolation
  • TPM2 unlockoddlama: Bypassing disk encryption on systems with automatic TPM2 unlock (write-up, Jan 2025)
  • TPM busCPU to TPM Bus Protection Guidance, Active Attack Mitigations (TCG, v1.0 r30)
  • CVE-2026-0714TPM-sniffing LUKS Keys on an Embedded Device (Cyloq, Feb 2026)
  • ThunderclapA. T. Markettos et al.: Exploring Vulnerabilities in Operating System IOMMU Protection via DMA from Untrustworthy Peripherals (NDSS, 2019)
  • JanuscapeCVE-2026-53359, Guest-to-Host Escape in KVM/x86 (nested virtualization) (oss-security, Jul 2026)
Cryptographic constructions & dispersed storage
  • AONT-RSJ. Resch, J. Plank: AONT-RS: Blending Security and Performance in Dispersed Storage Systems (USENIX FAST, 2011)
  • AONTR. L. Rivest: All-or-Nothing Encryption and the Package Transform (FSE, 1997)
WireGuard & post-quantum tunnels
  • WireGuardJ. A. Donenfeld: WireGuard, Next Generation Kernel Network Tunnel (NDSS, 2017)
  • PQ-WireGuardA. Hülsing, K.-C. Ning, P. Schwabe, F. Weber, P. Zimmermann: Post-quantum WireGuard (IACR ePrint 2020/379)
  • RosenpassRosenpass whitepaper: post-quantum key exchange for WireGuard (Classic McEliece, Kyber, ProVerif analysis) (Rosenpass project)
Consensus & quorum
  • PaxosL. Lamport: The Part-Time Parliament (ACM TOCS, 1998)
  • RaftD. Ongaro, J. Ousterhout: In Search of an Understandable Consensus Algorithm (USENIX ATC, 2014)

Thank you to the people who break things so the rest of us don't have to

Almost everything on this page stands on work that someone else did first. Researchers who spend nights and weekends with logic analyzers, voltage glitchers and debuggers, who write up what they found in public, and who then send a polite disclosure to a vendor and wait patiently for the answer. Standards authors who argue over a single sentence in an RFC for years so the rest of the internet can rely on it. Maintainers of open-source projects who answer bug reports for free.

You make the internet a safer place, mostly without recognition and often without payment. Our infrastructure is better because of you, and so are our customers. Thank you, sincerely, and keep breaking things responsibly.

HU.HN mascot: a stern white rooster with arms crossed and a gold chain, inside a red ring

Questions? Want the full picture?

A more detailed, non-public overview is available to our customers on request, after signing a strictly binding non-disclosure agreement.

team@hu.hn