Skip to content

NetSecOPEN + RFC 9411 alignment

Scope status (post-Scope-Freeze 2026-05-10) — See ARCHITECTURE.md for the canonical 37 MÓDULOs + 7 Test Kinds + DOM/CPOS/PIE-PA safety architecture. ADRs 0014, 0019-0025 cover post-Freeze additions.

Strategic positioning: TLSStress.Art is not a NetSecOPEN approved-tool (Spirent Avalanche / Keysight CyPerf-BreakingPoint are the only two as of 2026). We are however vocabulary-aligned + methodology-aligned with RFC 9411 + NetSecOPEN test plans, so a customer can cross-reference our reports against any NetSecOPEN-certified vendor datasheet without translation friction.

Per the locked decision in discuss_netsecopen_rfc9411.md (Option C: RFC 9411-compliant + Production Realism extensions).

Standards reference

  • RFC 9411Benchmarking Methodology for Network Security Device Performance (Informational, Mar 2023, obsoletes RFC 3511). Authors include NetSecOPEN consortium members.
  • RFC 2544 — Benchmarking Methodology for Network Interconnect Devices (older but still cited for L2/L3 reference points).
  • RFC 6349 — Framework for TCP Throughput Testing (used by our Branch Office tests for asymmetric WAN scenarios).
  • NetSecOPEN test methodology documents — published by the NetSecOPEN consortium; cite RFC 9411 + add per-vendor approved certification procedures.

Vocabulary mapping (TLSStress.Art ↔ RFC 9411)

TLSStress.Art term RFC 9411 § Notes
TLS Decryption / TLS Inspection §3.1.5 + §6.x We never use deprecated "SSL Inspection" — see TLS_INSPECTION_TRAPS.md
HTTP Throughput (Mbps / Gbps) §6.1.1 (HTTP Throughput Capacity) Reported per-quadrant in Annex I
HTTPS Throughput (Mbps / Gbps) §6.1.2 (HTTPS Throughput Capacity) Reported per-quadrant + per-cipher
Connections Per Second (CPS) §6.2 (Connection Rate) We split: HTTP-CPS, HTTPS-CPS, TLS-handshake-only
Concurrent Connections §6.3 (Concurrent Connection Capacity) Tracked via NAT/conntrack table at peak
Time To First Byte (TTFB) §6.4 (Latency / Application Response) p50, p95, p99, max — Annex H
Application Mix §6.5 (Application Transaction Capacity) Our Synthetic + Cloned Personas + HAR replay
Inspection Profile (5 named) §6 §6.1-§6.5 dial subset See INSPECTION_PROFILE.md §"RFC 9411 §6 cross-reference"
Quadrant Decomposition (Q1-Q4) n/a (TLSStress.Art-original) Adds production-realism beyond RFC 9411 — Annex I
Branch Office (asymmetric WAN) RFC 6349 (not RFC 9411) Annex J
BGP routing-table saturation n/a (TLSStress.Art-original) Control-plane stress; ADR 0012
Geo-CA distribution n/a (TLSStress.Art-original) Annex F (multi-region cert chain validation)
Airgap attestation n/a (TLSStress.Art-original) Annex G (5-layer L2/L3 isolation chain)

RFC 9411 §6 benchmarks — coverage matrix

RFC 9411 §6 benchmark TLSStress.Art coverage Where
§6.1.1 HTTP Throughput Capacity ✅ full browser-engine + synthetic-load fleets, Annex I
§6.1.2 HTTPS Throughput Capacity ✅ full TLS leg 2 to Caddy origins, Annex I
§6.2 Connection Rate ✅ full synthetic-load engine ramp + browser engine sustained
§6.3 Concurrent Connection Capacity ✅ full NAT/conntrack peak, Annex H
§6.4 Latency / Application Response ✅ full TTFB p50/p95/p99/max, Annex H
§6.5 Application Transaction Capacity ✅ partial Synthetic + cloned personas; HAR replay; does NOT include the proprietary app-replay library Spirent/Keysight ship (Office 365, Skype, Salesforce) — see Ixia comparativo
§6.6 (DDoS) Not in scope (we are not an attack-traffic generator)
§6.7 (Malware) Not in scope; future BreakingPoint-equivalent track

Production-realism extensions (NetSecOPEN-aligned, beyond RFC 9411)

These are TLSStress.Art-original extensions that map cleanly onto NetSecOPEN's "production realism" goals without violating any RFC 9411 invariant:

  1. Quadrant Decomposition (Q1-Q4) — partitions the run into four traffic-shape quadrants so a single number ("HTTPS throughput") is replaced with per-quadrant breakdowns. Useful when a vendor passes the aggregate but fails specific quadrants. Annex I.
  2. Inspection Profile — 5 named profiles (minimal, balanced, paranoid, compliance, sandbox) that map onto RFC 9411 §6 dial subsets. Per INSPECTION_PROFILE.md.
  3. Cloned Personas — real public sites cloned via headful browser engine into the bench. Closer fidelity than synthetic patterns; complements (does not replace) HAR replay.
  4. Geo-CA distribution — multi-region cert chain validation, relevant for global SD-WAN deployments. Annex F.
  5. Airgap attestation — 5-layer L2/L3 isolation proof. Annex G.

Annex labels (cross-reference quick lookup)

For operators reading our reports against a NetSecOPEN-certified vendor datasheet:

Annex TLSStress.Art topic RFC 9411 § anchored
F Geo-CA distribution n/a (original)
G Airgap attestation n/a (original)
H NAT engine performance §6.3 (concurrent) + §6.4 (latency)
I Quadrant decomposition §6.1.1 + §6.1.2 + §6.2 (per-quadrant)
J Branch Office WAN RFC 6349 (not RFC 9411)
K Inspection Profile (planned) §6 dial subset
L BGP routing-table saturation (planned) n/a (original)

What this is NOT

  • Not a NetSecOPEN approved-tool listing — we are not in the consortium's certification program.
  • Not a claim that our results are interchangeable with Spirent / Keysight numbers — apples-to-apples requires running both tools against the same DUT under identical conditions.
  • Not a pre-test of NetSecOPEN certification — vendors needing certification still go through Spirent / Keysight; we provide a sanity-check + production-realism overlay.

Future workstreams

  1. NetSecOPEN consortium membership (Option D in the original memo) — under separate strategic decision.
  2. §6.6 DDoS + §6.7 Malware — needs the attack/exploit content library gap identified in the Ixia comparativo.
  3. Per-Annex RFC 9411 §X labels in PDF report headers — small doc-only PR that adds the §X tag next to each annex title in the rendered report.