From 07b64a2a9e66a53c5c0af7725ac6f1f2de4249e3 Mon Sep 17 00:00:00 2001 From: leetcrypt Date: Thu, 23 Jul 2026 10:30:54 -0700 Subject: [PATCH] =?UTF-8?q?docs:=20honesty/accuracy=20pass=20=E2=80=94=20p?= =?UTF-8?q?hone-role=20disclosure,=20ICM=20provenance,=20dangling-citation?= =?UTF-8?q?=20fix?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit - Correct grid disclosure to match the sealed device-map: single-laptop isolated-docker host; the two phones are pinned consenting-node labels (can_host_engine=false / isolated_engine_available=false), never forwarders and carrying no measured traffic — across abstract, apparatus, scope, and limitations, plus the companion methods. - Replace two dangling `PHONE-ROLE-AUDIT.md` citations (file never existed) with the real artifact that substantiates the claim: grid/device-map.json; vendor that artifact so the citation resolves in a clean clone. - Drop the stale "(skeleton — quantitative claims held...)" abstract label now that the abstract is filled and RQ2 is ratified. - Disclose that the study was staged, pre-registered, and executed under the Interpretable Context Methodology (ICM) [VanClief2026]; add the reference. - Repoint the bibliography citation to the vendored docs/sor-consent-bibliography.md. - Regenerate the paper site from the corrected markdown. Co-Authored-By: Claude Opus 4.6 --- docs/paper-site/index.html | 37 ++++++++++---- docs/stage-07-companion-methods.md | 2 +- docs/stage-07-paper-draft.md | 33 ++++++++---- .../20260720T060132Z/grid/device-map.json | 51 +++++++++++++++++++ 4 files changed, 102 insertions(+), 21 deletions(-) create mode 100644 output/sor-confirmatory/20260720T060132Z/grid/device-map.json diff --git a/docs/paper-site/index.html b/docs/paper-site/index.html index b9cfce5..bc5c120 100644 --- a/docs/paper-site/index.html +++ b/docs/paper-site/index.html @@ -569,7 +569,7 @@ anonymity set rather than growing it (RQ2-P1, a Holm-significant negative
-

Abstract (skeleton — quantitative claims held until data + RQ2 ratification)

+

Abstract

Onion-routing systems typically admit any relay that meets a directory's technical criteria; they do not model relay consent — a host's in-band, per-circuit choice to carry a given flow. We build and measure a consent-gated, federated, nested-SSH relay data plane in which @@ -577,7 +577,7 @@ every hop must explicitly accept or reject each circuit through a signed in-band (Ed25519-authenticated, X25519 per-hop credentials), and in which relays are organized into houses that federate either through a shared bridge or through a directory. Treating this as a measurement instrument for a trust model's exposure (not a service that provides -anonymity), we ask two confirmatory questions on a lab grid of two phones and a laptop: (RQ1) +anonymity), we ask two confirmatory questions on a single-laptop isolated-docker grid (two non-forwarding phones pinned): (RQ1) does a shared bridge introduce a measurable flow-linkability leak between a circuit's entry and exit segments, and does cover padding remove it; (RQ2) does federating relays across houses grow or shrink the anonymity set an adversary faces, and is any effect explained by @@ -635,8 +635,7 @@ linkability (RQ1) and the anonymity-set effect of federation (RQ2) on a lab grid when consent-gated federation helps or harms anonymity. On this instrument the answer is a double null/negative: no bridge leak to close, and federation that measurably reduces the anonymity set — reported here without spin as the paper's evidentiary core.

-

Scope. Claims are deliberately restricted to the tested lab topology and scale (two phones + -laptop, few houses); this is not an internet-scale or global-passive-adversary result (§7). +

Scope. Claims are deliberately restricted to the tested lab topology and scale (a single laptop's isolated-docker containers; two phones pinned but non-forwarding; few houses); this is not an internet-scale or global-passive-adversary result (§7). The paired churn-resilience question (RQ3) and a QUIC/ssh3 transport arm [Michel2023] are pre-registered but held for a companion paper; this lead paper covers G4 + RQ1 + RQ2 only.

@@ -700,7 +699,14 @@ by the freeze. Key mechanisms:

4. Methods (pre-registered; frozen)

This study is a confirmatory factorial controlled comparison; the design, variables, seeds, -detectors, and analysis were frozen and hashed on 2026-07-19 before any confirmatory cell ran.

+detectors, and analysis were frozen and hashed on 2026-07-19 before any confirmatory cell ran. +It was designed, pre-registered, and executed under the Interpretable Context Methodology +(ICM) [VanClief2026], a staged-pipeline framework in which each phase — literature, hypothesis, +design/pre-registration, build, execution, analysis, and write-up — is a numbered stage whose +frozen output/ is the sole input to the next. ICM is the structural mechanism behind the +freeze-before-data discipline used throughout this section: the pre-registration was frozen and +SHA-256-sealed in the design stage before the build and execution stages could consume it, so the +provenance chain (§4.4) is auditable by construction rather than by convention.

4.1 Design matrix

Cells are organised per RQ with the other factors held at their declared control:

    @@ -744,7 +750,10 @@ completion — no optional stopping, no interim looks; an uninf inconclusive, never extended to chase significance.

    Apparatus (disclosed). All relay hops ran as isolated Docker containers on a single engine host (the laptop; grid/device-map.json, isolated_engine_host_count = 1, Docker 27.5.1). The -two phones were consenting endpoints, not forwarders. Node distinctness is thus container-level +two phones were pinned consenting-node labels, not forwarders (probed reachable once at grid-pin; +carried no measured traffic — the device map records both phones with can_host_engine = false / +isolated_engine_available = false, i.e. structurally unable to run an isolated forwarder). Node +distinctness is thus container-level (≥ 3 distinct containers per circuit), and matched-N is pinned from the containerised node count per manifest; cross-machine effects are out of scope (§7).

    4.4 Frozen detectors and the instrument-validation gate

    @@ -940,14 +949,17 @@ without being explained away.

    7. Limitations & threats to validity

      -
    • Scale / adversary model (External). The grid is two phones + a laptop and few houses; this +
    • Scale / adversary model (External). The grid is a single laptop (isolated-docker) with two non-forwarding phones pinned, and few houses; this is not internet-scale and not a global passive adversary. Claims are scoped to the tested topology/scale; entropy CIs are wide at small node counts (accepted, node counts reported).
    • Node distribution (External, disclosed). All relay hops executed as isolated Docker containers on a single engine host (the laptop; isolated_engine_host_count = 1, recorded in - grid/device-map.json). The two phones were consenting endpoints, not forwarders — they - cannot host an isolated engine. Node distinctness for RQ1/RQ2 is therefore container-level + grid/device-map.json). The two phones were pinned consenting-node labels, not forwarders — they + cannot host an isolated engine (grid/device-map.json records both phones with + can_host_engine = false / isolated_engine_available = false), were verified reachable only by a + single grid-pin probe, and carried no measured traffic. Node distinctness for RQ1/RQ2 is therefore + container-level (≥ 3 distinct containers per circuit), not physical-machine-level; matched-N is pinned from the containerised node count per manifest. This satisfies the containment law (every forwarder runs in an isolated engine, engine ≠ local) but means cross-machine timing effects are out of @@ -1032,8 +1044,11 @@ estimate, CI gate, or decision is substituted. The frozen prereg SHA is unchange
    • [Mittal2012b] Mittal, P., Caesar, M., & Borisov, N. (2012). X-Vine. NDSS 2012. arXiv:1109.0971.
    • [Zhou2011] Zhou, P., et al. (2011/2013). STor. arXiv:1110.5794.
      • +
    • [VanClief2026] Van Clief, J., & McDermott, D. (2026). Interpretable Context Methodology: + Folder Structure as Agentic Architecture. arXiv:2603.16021. (Methodology framework under which + this study was staged, pre-registered, and executed.)
    -

    (Full bibliography: ~/coding/sci-method/stages/01-literature/output/sor-consent-bibliography.md. +

    (Full bibliography: docs/sor-consent-bibliography.md (vendored in-repo). Integrity flags carried forward: [Stutzbach2006] secondary-sourced; [Constantinides2026] recent preprint — neither is load-bearing in this lead paper.)

    @@ -1383,7 +1398,7 @@ confirmatory battery and have both passed on a dry, synthetic, offlinecaught the unique-bridge artifact and promotes the corrected mechanism finding into the surviving family. Lead RQ1-P1 and RQ2-P1 survive regardless.
  • Scope & limitations. Both findings are scoped to the lab grid (1-house / bridge-off control, - 2 phones + laptop, self-generated fixture traffic) and inherit the lead paper's external-validity + single-laptop isolated-docker, two non-forwarding phones, self-generated fixture traffic) and inherit the lead paper's external-validity caveats. Specific to this companion: (i) the RQ2-P3 mix is an as-instrumented concentration effect on the ratified exit-signature posterior, not an internet-scale claim; (ii) the RQ3 nulls are grid-bound — the perf null follows from a baseline retention ceiling under the pinned diff --git a/docs/stage-07-companion-methods.md b/docs/stage-07-companion-methods.md index b90a5b0..db84cc4 100644 --- a/docs/stage-07-companion-methods.md +++ b/docs/stage-07-companion-methods.md @@ -343,7 +343,7 @@ confirmatory battery and have both passed on a **dry, synthetic, offline** pass: companion's frozen-detector method both *caught* the unique-bridge artifact and *promotes* the corrected mechanism finding into the surviving family. Lead RQ1-P1 and RQ2-P1 survive regardless. - **Scope & limitations.** Both findings are scoped to the lab grid (1-house / bridge-off control, - 2 phones + laptop, self-generated fixture traffic) and inherit the lead paper's external-validity + single-laptop isolated-docker, two non-forwarding phones, self-generated fixture traffic) and inherit the lead paper's external-validity caveats. Specific to this companion: (i) the RQ2-P3 mix is an **as-instrumented** concentration effect on the ratified exit-signature posterior, not an internet-scale claim; (ii) the RQ3 nulls are **grid-bound** — the perf null follows from a baseline retention ceiling under the pinned diff --git a/docs/stage-07-paper-draft.md b/docs/stage-07-paper-draft.md index 062ce51..0c8db48 100644 --- a/docs/stage-07-paper-draft.md +++ b/docs/stage-07-paper-draft.md @@ -26,7 +26,7 @@ --- -## Abstract *(skeleton — quantitative claims held until data + RQ2 ratification)* +## Abstract Onion-routing systems typically admit any relay that meets a directory's technical criteria; they do not model **relay consent** — a host's in-band, per-circuit choice to carry a given @@ -35,7 +35,7 @@ every hop must explicitly accept or reject each circuit through a signed in-band (Ed25519-authenticated, X25519 per-hop credentials), and in which relays are organized into **houses** that federate either through a shared **bridge** or through a **directory**. Treating this as a *measurement instrument* for a trust model's exposure (not a service that provides -anonymity), we ask two confirmatory questions on a lab grid of two phones and a laptop: **(RQ1)** +anonymity), we ask two confirmatory questions on a single-laptop isolated-docker grid (two non-forwarding phones pinned): **(RQ1)** does a shared bridge introduce a measurable flow-linkability leak between a circuit's entry and exit segments, and does cover padding remove it; **(RQ2)** does federating relays across houses **grow or shrink** the anonymity set an adversary faces, and is any effect explained by @@ -95,8 +95,7 @@ when consent-gated federation helps or harms anonymity. **On this instrument the double null/negative: no bridge leak to close, and federation that measurably *reduces* the anonymity set** — reported here without spin as the paper's evidentiary core. -**Scope.** Claims are deliberately restricted to the tested lab topology and scale (two phones + -laptop, few houses); this is not an internet-scale or global-passive-adversary result (§7). +**Scope.** Claims are deliberately restricted to the tested lab topology and scale (a single laptop's isolated-docker containers; two phones pinned but non-forwarding; few houses); this is not an internet-scale or global-passive-adversary result (§7). The paired **churn-resilience** question (RQ3) and a QUIC/`ssh3` transport arm [Michel2023] are pre-registered but held for a companion paper; this lead paper covers G4 + RQ1 + RQ2 only. @@ -172,6 +171,13 @@ by the freeze. Key mechanisms: This study is a **confirmatory factorial controlled comparison**; the design, variables, seeds, detectors, and analysis were frozen and hashed on 2026-07-19 before any confirmatory cell ran. +It was designed, pre-registered, and executed under the **Interpretable Context Methodology** +(ICM) [VanClief2026], a staged-pipeline framework in which each phase — literature, hypothesis, +design/pre-registration, build, execution, analysis, and write-up — is a numbered stage whose +frozen `output/` is the sole input to the next. ICM is the structural mechanism behind the +freeze-before-data discipline used throughout this section: the pre-registration was frozen and +SHA-256-sealed in the design stage before the build and execution stages could consume it, so the +provenance chain (§4.4) is auditable by construction rather than by convention. ### 4.1 Design matrix @@ -218,7 +224,10 @@ completion — **no optional stopping, no interim looks**; an uninformative cell **Apparatus (disclosed).** All relay hops ran as **isolated Docker containers on a single engine host** (the laptop; `grid/device-map.json`, `isolated_engine_host_count = 1`, Docker 27.5.1). The -two phones were **consenting endpoints, not forwarders**. Node distinctness is thus container-level +two phones were **pinned consenting-node labels, not forwarders** (probed reachable once at grid-pin; +carried no measured traffic — the device map records both phones with `can_host_engine = false` / +`isolated_engine_available = false`, i.e. structurally unable to run an isolated forwarder). Node +distinctness is thus container-level (≥ 3 distinct containers per circuit), and matched-N is pinned from the containerised node count per manifest; cross-machine effects are out of scope (§7). @@ -391,14 +400,17 @@ without being explained away. ## 7. Limitations & threats to validity -- **Scale / adversary model (External).** The grid is two phones + a laptop and few houses; this +- **Scale / adversary model (External).** The grid is a single laptop (isolated-docker) with two non-forwarding phones pinned, and few houses; this is **not** internet-scale and **not** a global passive adversary. Claims are scoped to the tested topology/scale; entropy CIs are wide at small node counts (accepted, node counts reported). - **Node distribution (External, disclosed).** All relay hops executed as **isolated Docker containers on a single engine host** (the laptop; `isolated_engine_host_count = 1`, recorded in - `grid/device-map.json`). The two phones were **consenting endpoints, not forwarders** — they - cannot host an isolated engine. Node *distinctness* for RQ1/RQ2 is therefore container-level + `grid/device-map.json`). The two phones were **pinned consenting-node labels, not forwarders** — they + cannot host an isolated engine (`grid/device-map.json` records both phones with + `can_host_engine = false` / `isolated_engine_available = false`), were verified reachable only by a + single grid-pin probe, and carried no measured traffic. Node *distinctness* for RQ1/RQ2 is therefore + container-level (≥ 3 distinct containers per circuit), not physical-machine-level; matched-N is pinned from the containerised node count per manifest. This satisfies the containment law (every forwarder runs in an isolated engine, `engine ≠ local`) but means cross-machine timing effects are **out of @@ -487,7 +499,10 @@ estimate, CI gate, or decision is substituted. The frozen prereg SHA is unchange - **[Mittal2012b]** Mittal, P., Caesar, M., & Borisov, N. (2012). X-Vine. *NDSS 2012*. arXiv:1109.0971. - **[Zhou2011]** Zhou, P., et al. (2011/2013). STor. arXiv:1110.5794. +- **[VanClief2026]** Van Clief, J., & McDermott, D. (2026). Interpretable Context Methodology: + Folder Structure as Agentic Architecture. arXiv:2603.16021. *(Methodology framework under which + this study was staged, pre-registered, and executed.)* -*(Full bibliography: `~/coding/sci-method/stages/01-literature/output/sor-consent-bibliography.md`. +*(Full bibliography: `docs/sor-consent-bibliography.md` (vendored in-repo). Integrity flags carried forward: [Stutzbach2006] secondary-sourced; [Constantinides2026] recent preprint — neither is load-bearing in this lead paper.)* diff --git a/output/sor-confirmatory/20260720T060132Z/grid/device-map.json b/output/sor-confirmatory/20260720T060132Z/grid/device-map.json new file mode 100644 index 0000000..1bfb5be --- /dev/null +++ b/output/sor-confirmatory/20260720T060132Z/grid/device-map.json @@ -0,0 +1,51 @@ +{ + "assessment": "GO on isolated docker host", + "containment": { + "external_target": "none", + "hops_run_in": "isolated docker containers only (never a phone/host forwarder)", + "live_vm_churn_on_rq1_rq2": "none (RQ1/RQ2 use no churn; churn_schedule_id=none)", + "self_traffic_only": true + }, + "degraded": false, + "devices": [ + { + "arch": "x86_64", + "can_host_engine": true, + "engine_version": "27.5.1", + "house_role": "house-A docker host + hop pool", + "isolated_engine_available": true, + "kind": "laptop", + "name": "laptop", + "ssh_reachable": true + }, + { + "arch": "aarch64", + "can_host_engine": false, + "engine_version": null, + "house_role": "house-B consenting node (phone)", + "isolated_engine_available": false, + "kind": "phone", + "name": "fp6", + "ssh_reachable": true + }, + { + "arch": "aarch64", + "can_host_engine": false, + "engine_version": null, + "house_role": "house-C consenting node (Termux, no docker)", + "isolated_engine_available": false, + "kind": "phone", + "name": "tril", + "ssh_reachable": true + } + ], + "devices_down": [], + "generated_utc": "2026-07-20T06:01:52Z", + "isolated_engine_host_count": 1, + "local_docker_version": "27.5.1", + "matched_N_note": "RQ1/RQ2 isolated hops are containerised on the docker host (>=3 distinct containers = distinct nodes). Physical-phone distribution is optional; the confirmatory matched-N (single-house N = federated total consenting nodes) is pinned from the containerised node count at run time and recorded per manifest.", + "reachable_count": 3, + "schema": "sor-device-map/1", + "scope": "RQ1+RQ2 lead-paper grid pin (CLAUDE.md \u00a7Containment)", + "topology_matchedN_honourable": true +}