Join our Newsletter — 33% off our NHI Course
Home› Guides› TLS Server Configuration Baseline Guide: Versions, Cipher Suites,…
Guide Network Security

TLS Server Configuration Baseline Guide: Versions, Cipher Suites, HSTS and Testing in One Standard

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 27 October 2026 13 min read
On this page

Most organisations run TLS on hundreds of endpoints, from web servers and load balancers to APIs, mail gateways and appliances, and each one negotiates whatever its defaults allow unless someone sets and checks a single baseline. The two sources that define that baseline are NIST SP 800-52 Rev. 2, which sets TLS requirements for US government systems, and RFC 9325 (BCP 195), the IETF's current recommendations for secure use of TLS. Neither tells you exactly what to type into a load balancer, so this guide turns them into one enforceable server baseline: protocol versions, cipher suites, key exchange groups, certificates, HSTS, session handling and revocation, with implementation steps, test evidence and framework mapping. It is written for platform, network and application security engineers and for auditors. How the protocol works is covered in the How TLS Works Guide.

Key takeaways

  • TLS 1.2 and 1.3 only. RFC 8996 says TLS 1.0 and 1.1 MUST NOT be used; RFC 9325 makes TLS 1.2 mandatory and TLS 1.3 the preferred version; SP 800-52 Rev. 2 required TLS 1.3 support in government systems by 1 January 2024.
  • Forward secrecy and AEAD only for TLS 1.2. The four ECDHE AES-GCM suites in RFC 9325, optionally plus ChaCha20-Poly1305; no static RSA, no CBC without encrypt_then_mac, nothing below 128 bits.
  • Keys and signatures have floors. RSA at least 2048 bits, ECDHE curves of at least 224 bits (use X25519 and P-256), DH at least 2048 bits, and no SHA-1 or MD5 signatures.
  • HSTS on every web host. RFC 9325 says web servers SHOULD use HSTS; set a long max-age once every subdomain is ready for HTTPS.
  • Defaults lie. Some platforms still default to TLS 1.0 for API-created listeners, so scan every endpoint externally and treat the scan output as the evidence.

What it is and who it applies to

SourceStatus (checked 27 October 2026)Applies toRole in the baseline
NIST SP 800-52 Rev. 2Final, 29 August 2019; NIST notes (May 2026) it is under review through its Crypto Publication Review ProjectUS federal systems; widely used as a reference elsewhereRequires TLS 1.2 with FIPS-based cipher suites on all government servers and clients, and TLS 1.3 support by 1 January 2024
RFC 9325 (BCP 195)Best Current Practice, November 2022; obsoletes RFC 7525Any TLS or DTLS deploymentThe detailed MUST and SHOULD rules for versions, suites, keys, resumption and extensions
RFC 8996 (BCP 195)Best Current Practice, March 2021Any TLS deploymentFormally deprecates TLS 1.0, TLS 1.1 and DTLS 1.0
RFC 6797 (HSTS)Standards Track, November 2012HTTPS web hostsDefines Strict-Transport-Security
Server Side TLS guidelines (tlsref, formerly Mozilla)Version 6.0Web servers and proxiesPractical "intermediate" and "modern" profiles with exact suite lists

Scope: every TLS server you operate: web servers, reverse proxies, load balancers and CDNs, API gateways, mail servers (see the SMTP Security, STARTTLS and MTA-STS Guide), databases, message brokers, management interfaces of appliances and internal services. Internal endpoints need the same baseline; attackers inside the network can intercept traffic too (ATT&CK T1557 Adversary-in-the-Middle, T1040 Network Sniffing).

Structure and key requirements

Protocol versions and handshake behaviour

SettingBaselineSource
SSL 2.0, SSL 3.0, TLS 1.0, TLS 1.1DisabledRFC 8996; RFC 9325 3.1.1
TLS 1.2EnabledRFC 9325 3.1.1 (MUST); SP 800-52 Rev. 2
TLS 1.3Enabled and preferredRFC 9325 3.1.1 (SHOULD); SP 800-52 Rev. 2 (support by 1 January 2024)
Version fallbackNone below TLS 1.2RFC 9325 3.1.3
TLS compressionOffRFC 9325 3.3
Renegotiation (TLS 1.2)Secure renegotiation (renegotiation_info) required; client-initiated renegotiation off where the server allowsRFC 9325 3.5
Extended master secret (TLS 1.2)SupportedRFC 9325 3.5
SNI and ALPNSupported; reject unknown names; enforce ALPNRFC 9325 3.7 and 3.8
0-RTT early data (TLS 1.3)Off unless the application protocol explicitly allows itRFC 9325 3.10
Truncated HMACOffRFC 9325 4.6

Cipher suites and key exchange

ProtocolAllowedNot allowed
TLS 1.3TLS_AES_128_GCM_SHA256, TLS_AES_256_GCM_SHA384, TLS_CHACHA20_POLY1305_SHA256 (the tlsref profiles enable these three)CCM_8 suites outside constrained-device profiles
TLS 1.2ECDHE-ECDSA and ECDHE-RSA with AES_128_GCM_SHA256 and AES_256_GCM_SHA384 (RFC 9325 4.2); the tlsref intermediate profile adds ECDHE CHACHA20-POLY1305NULL, RC4, export and anything below 112 bits (MUST NOT); below 128 bits, static RSA key transport, static DH or ECDH, and DHE (SHOULD NOT); CBC suites unless encrypt_then_mac is negotiated
GroupsX25519 and P-256 (RFC 9325 4.2.1); tlsref 6.0 lists X25519MLKEM768, X25519, prime256v1 and secp384r1Curves below 224 bits; DH groups below 2048 bits

Server cipher order matters less than it used to: the tlsref profiles let the client choose, because every allowed suite is strong. FIPS-constrained systems should drop ChaCha20-Poly1305 and X25519 if their validated module does not include them. Hybrid post-quantum groups such as X25519MLKEM768 can be added without harm to clients that do not support them; see the Hybrid Post-Quantum TLS Key Exchange Guide.

Certificates, keys, sessions and HSTS

  • Keys: ECDSA P-256 (the tlsref recommendation) or RSA of at least 2048 bits (RFC 9325 4.5). SHA-256 or stronger signatures; SHA-1 and MD5 MUST NOT be used.
  • Chain: serve the full chain without the root, with the leaf's subjectAltName covering every hostname. Short-lived certificates (tlsref recommends 90 days) need automation; see the ACME and Automated Certificate Management Guide and the Certificate Expiry Outages and 47-Day TLS Certificates Guide.
  • Revocation: RFC 9325 7.5 says implementations MUST have a strategy to distrust revoked certificates and servers SHOULD support OCSP stapling. Check whether your issuing CA still runs OCSP before enabling stapling; see the Certificate Revocation Guide.
  • Session resumption: tickets must be encrypted and authenticated with a cipher at least as strong as the connection's, ticket keys rotated regularly (RFC 9325 gives weekly as an example) and old keys destroyed, and ticket lifetime limited (RFC 9325 3.4). Across a load-balanced pool, distribute and rotate ticket keys centrally or disable tickets.
  • HSTS: RFC 6797 defines max-age (required) and includeSubDomains (optional), and the header is only honoured over HTTPS. The tlsref profiles use max-age=63072000 (two years). The preload list is a browser programme, not part of the RFC; only submit once every subdomain serves HTTPS, because removal is slow.
  • TLS-only ports: where a protocol offers both an implicit TLS port and a STARTTLS upgrade, RFC 9325 3.2 says administrators must prefer the TLS-only port.

Choosing a profile

SituationProfileWhy
Public website with unknown browsersStandard (TLS 1.2 and 1.3, intermediate suites)tlsref lists oldest supported clients such as Android 4.4.2, Chrome 49 and Java 8u161, covering almost all real traffic
New API or service whose clients you controlStrict (TLS 1.3 only)tlsref's modern profile needs clients from roughly Chrome 70, Java 11 and OpenSSL 1.1.1 onwards and removes all TLS 1.2 choices
US federal or FIPS-bound systemStandard or strict, restricted to FIPS-approved suites and groupsSP 800-52 Rev. 2 requires FIPS-based cipher suites
Legacy clients that cannot be upgradedException with a separate endpoint and end dateKeeps weak settings off the main service, so the exception can be removed cleanly

Small vs large organisations: a small organisation using a managed CDN or cloud load balancer can meet most of the baseline by choosing the provider's TLS 1.2-minimum or TLS 1.3 policy, turning on HSTS and scanning monthly. A large organisation needs the baseline written as code, per-platform templates, a scanning service that covers internal ranges, and an exception process owned by security architecture.

Identity and NHI requirements

Every TLS server certificate is a machine identity, and its private key is a credential. A stolen key lets an attacker impersonate the service or, without forward secrecy, decrypt recorded TLS 1.2 traffic that used static RSA key exchange, which is one reason the baseline removes static RSA suites.

  • Key custody: generate keys on the server, load balancer or HSM; never email or commit them. Restrict file permissions and cloud certificate store access to the service and its automation identity.
  • Automation identities: ACME clients, certificate manager integrations and infrastructure as code pipelines that can issue or deploy certificates hold powerful rights; scope and rotate their credentials like any other non-human identity. See the Machine Identity, PKI and Certificate Lifecycle Guide.
  • Inventory: keep every certificate, its owner, its key location and its expiry in one inventory, fed by scans and Certificate Transparency monitoring.
  • Client authentication: where services authenticate each other, apply the same baseline to mutual TLS and validate client certificates properly. See the mTLS Guide.
  • Ticket keys: session ticket encryption keys are secrets shared across a server pool; store and rotate them in a secrets manager.

Implementation steps

  1. Discover: scan all external and internal address ranges for TLS on any port and build an inventory of endpoints, software, certificate and current protocol support. Include appliances and management interfaces.
  2. Write the baseline: adopt the tables above as a configuration standard with two profiles: "standard" (TLS 1.2 and 1.3, the intermediate suite list) and "strict" (TLS 1.3 only) for new services and APIs whose clients you control.
  3. Translate per platform: nginx (ssl_protocols, ssl_ciphers, ssl_conf_command for TLS 1.3 suites), Apache (SSLProtocol, SSLCipherSuite), Windows Schannel (the Enabled DWORD under HKLM\SYSTEM\CurrentControlSet\Control\SecurityProviders\SCHANNEL\Protocols\<version>\Server, with cipher suite order set through Group Policy, MDM or PowerShell rather than direct registry edits, as Microsoft advises), and cloud load balancer security policies.
  4. Check cloud defaults: on AWS Application Load Balancers, the console default is ELBSecurityPolicy-TLS13-1-2-Res-PQ-2025-09, but listeners created through the CLI, CloudFormation or CDK default to ELBSecurityPolicy-2016-08, which still allows TLS 1.0 and 1.1. Set the policy explicitly in every template.
  5. Measure client impact first: before disabling TLS 1.0 and 1.1 or CBC suites, log the negotiated version and suite per connection for two to four weeks to find legacy clients. Contact owners, then disable on a pilot group of endpoints.
  6. Roll out in waves: internet-facing web first, then APIs, then internal services and appliances. Rollback is restoring the previous configuration or security policy, which takes effect for new connections; on Windows a restart may be needed because Schannel settings apply to credential handles acquired after the change.
  7. Add HSTS gradually: start with a short max-age, then raise it to one or two years and add includeSubDomains once every subdomain serves HTTPS.
  8. Enforce continuously: scan weekly and on every deployment, and fail infrastructure as code checks that reference non-compliant policies.

Evidence and assessment

  • External scan output: results from a TLS scanner (such as testssl.sh, SSLyze or the Nmap ssl-enum-ciphers script) for every endpoint, showing protocols, suites, groups, certificate details and HSTS. This is stronger evidence than configuration files, because it shows what is actually negotiated.
  • Configuration evidence: the baseline document, per-platform templates and the load balancer security policy attached to each listener.
  • Exception register: each endpoint allowed to deviate, the client population that needs it, the compensating control and the end date.
  • Certificate inventory: key type and size, issuer, expiry, owner and automation status.
MetricTarget
Endpoints negotiating TLS 1.0 or 1.1Zero outside the exception register
Endpoints offering any suite outside the baselineZero outside the exception register
Endpoints with TLS 1.3 enabled100% of platforms that support it
Web hosts sending HSTS with max-age of at least one year100% of HTTPS web hosts
Certificates with RSA below 2048 bits or SHA-1 signaturesZero
Endpoints scanned in the last seven days100% of inventory

Mapping to other frameworks

Baseline elementNIST SP 800-53 Rev. 5NIST CSF 2.0CIS Controls v8.1ISO/IEC 27001:2022
Protocol versions and cipher suitesSC-8, SC-8(1), SC-13PR.DS-023.10A.8.24
Certificates and keysSC-17, SC-12PR.AA-033.10A.8.24
Session handling, HSTS, renegotiationSC-23PR.DS-0216A.8.26
Configuration baseline and enforcementCM-2, CM-6PR.PS-014.1A.8.9
Continuous scanningRA-5ID.RA-017A.8.8

Common pitfalls

  • Template defaults: infrastructure as code that omits the load balancer security policy and silently inherits a legacy default.
  • Front door only: the CDN or load balancer is hardened while the origin, admin interfaces and internal services still accept TLS 1.0 and weak suites.
  • Static RSA left in: TLS_RSA suites kept "for compatibility", removing forward secrecy for any client that picks them.
  • HSTS preload too early: preload submitted before every subdomain supports HTTPS, breaking internal or legacy hosts for months.
  • Ticket keys never rotated: one session ticket key shared across a pool for years, undermining forward secrecy for resumed sessions.
  • Configuration without verification: settings applied, but an older library or an appliance firmware ignores them; only an external scan shows the truth.

Practitioner checklist

  • Inventory every TLS endpoint, including internal services, appliances and management interfaces.
  • Disable SSL 2.0, SSL 3.0, TLS 1.0 and TLS 1.1; enable TLS 1.2 and TLS 1.3 everywhere.
  • Restrict TLS 1.2 to ECDHE with AES-GCM (and optionally ChaCha20-Poly1305); remove static RSA, DHE and CBC suites.
  • Use X25519 and P-256 groups, ECDSA P-256 or RSA 2048-bit certificates, and SHA-256 or stronger signatures.
  • Turn off compression, insecure renegotiation, truncated HMAC and 0-RTT unless explicitly required.
  • Rotate session ticket keys at least weekly and limit ticket lifetime, or disable tickets.
  • Send HSTS with a long max-age on every HTTPS web host, adding includeSubDomains when ready.
  • Set load balancer and CDN security policies explicitly in code; never rely on defaults.
  • Scan every endpoint weekly and on change, and keep an exception register with end dates.

Standards and references

Related NHI Mgmt Group resources: How TLS Works Guide · Certificate Expiry Outages and 47-Day TLS Certificates Guide · mTLS Guide · TLS Inspection Guide

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 27 October 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.