audit: add LEP-6 shadow scoring for storage truth#119
Closed
j-rafique wants to merge 3 commits intoLEP-6-report-ingestionfrom
Closed
audit: add LEP-6 shadow scoring for storage truth#119j-rafique wants to merge 3 commits intoLEP-6-report-ingestionfrom
j-rafique wants to merge 3 commits intoLEP-6-report-ingestionfrom
Conversation
Both previously flagged issues have been addressed in this revision. Params are now threaded through as parameters (no redundant store reads), and
Mention @roomote in a comment to request specific changes to this pull request or fix all unresolved issues. |
![roomote-v0[bot]](https://avatars.githubusercontent.com/in/1546624?s=60&v=4){kind=small}
j-rafique
force-pushed
the
LEP-6-shadow-scoring
branch
from
51ea7b0 to
c86ef0e
Compare
Contributor
Production-gate review by Zee — 12 findingsMethodology: full file-by-file read of every non-generated changed file in this PR's diff (pr-119 vs its base branch), cross-checked against:
Status legend: each finding's status is computed at the PR #122 stack-tip (consensus-gap-fixes commit Severity breakdown: HIGH=7, MEDIUM=5 119-F1 — Decay was linear instead of geometric (spec §14/§15/§16)
119-F2 — Node-suspicion deltas don't match spec §14 (no symbol/index split, wrong magnitudes)
119-F3 — Ticket-deterioration deltas don't match spec §16
119-F4 — Reporter-reliability deltas semantically inverted vs spec §15
119-F5 — Contradiction penalty applied to BOTH the contradicting and contradicted reporter
119-F6 — Trust multiplier 0.25 (LOW_TRUST=0.5, INELIGIBLE=0.25) violates spec §15.4 floor 0.5
119-F9 — Pattern-escalation predicate counts same-ticket repeats instead of distinct-ticket per node, magnitudes 2/4/6 vs spec 10/15/12
119-F10 —
|
Contributor
There was a problem hiding this comment.
Pull request overview
Implements LEP-6 PR3 “shadow scoring” for storage-truth by executing scoring on ingested storage_proof_results, persisting additional scoring/bookkeeping state, and emitting a dedicated score-update event.
Changes:
- Add storage-truth scoring pipeline on
MsgSubmitEpochReportwith node/reporter/ticket score deltas, trust-band scaling, decay-on-update, and contradiction/repeated-failure bookkeeping. - Extend protobuf state with
ReporterTrustBandand additional persisted fields for reporter and ticket scoring/bookkeeping. - Add/expand tests to cover scoring outcomes, decay, trust thresholds, escalation/contradictions, query visibility, and event attributes.
Reviewed changes
Copilot reviewed 10 out of 10 changed files in this pull request and generated 1 comment.
Show a summary per file
| File | Description |
|---|---|
| x/audit/v1/types/events.go | Introduces storage-truth score update event type and attribute keys. |
| x/audit/v1/types/audit.pb.go | Regenerated protobuf output to include trust band + added state fields. |
| x/audit/v1/keeper/storage_truth_scoring.go | Implements storage-truth scoring engine, bookkeeping, decay helpers, and event emission. |
| x/audit/v1/keeper/msg_submit_epoch_report.go | Hooks scoring execution into epoch report submission flow. |
| proto/lumera/audit/v1/audit.proto | Adds ReporterTrustBand enum and new persisted bookkeeping fields. |
| app/proto_bridge.go | Registers the new enum for proto bridge usage. |
| x/audit/v1/keeper/msg_submit_epoch_report_storage_truth_scores_test.go | Adds end-to-end tests for scoring, decay, scaling, events, escalation, contradictions. |
| x/audit/v1/keeper/query_storage_truth_test.go | Expands query tests to validate new fields and scored-ingestion visibility. |
| x/audit/v1/keeper/storage_truth_state_test.go | Updates state round-trip tests for newly persisted fields. |
| x/audit/v1/keeper/storage_truth_scoring_internal_test.go | Adds focused unit tests for internal math/scoring helpers. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+251
to
+256
| if isStorageTruthFailureClass(result.ResultClass) && epochID != state.LastFailureEpoch { | ||
| nextState.LastFailureEpoch = epochID | ||
| nextState.RecentFailureEpochCount = updateRecentFailureEpochCount(state, epochID, params) | ||
| } else if !found { | ||
| nextState.RecentFailureEpochCount = 0 | ||
| } |
There was a problem hiding this comment.
In applyTicketDeteriorationDelta, a brand-new ticket failure at epoch 0 won’t set LastFailureEpoch/RecentFailureEpochCount because the failure update is gated by epochID != state.LastFailureEpoch (and state.LastFailureEpoch is 0 for a new/zero-value state). This causes the first failure in epoch 0 to be recorded with RecentFailureEpochCount=0, and the next epoch’s failure will be treated as count=1 (no escalation) rather than count=2.
Consider changing the condition to also treat "not found" as needing initialization (e.g., update failure fields when !found || epochID != state.LastFailureEpoch), and add a test covering the first-ever failure when epochID == 0 to ensure RecentFailureEpochCount becomes 1.
j-rafique
force-pushed
the
LEP-6-report-ingestion
branch
from
cb16594 to
af4c9fc
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Implements LEP-6 PR3 (shadow scoring) on
LEP-6-shadow-scoring(PR #119) by adding storage-truth score execution for ingestedstorage_proof_resultsand persisting shadow-only scoring state.This PR updates state + scoring + observability only. It does not enable enforcement penalties.
Base / Scope
LEP-6-report-ingestionLEP-6-shadow-scoringSHADOW/SOFT/FULL, with PR3 intent focused on shadow bookkeeping and score evolution.What Changed
1) Storage-truth score pipeline execution
MsgSubmitEpochReportnow applies storage-truth scoring after report persistence.N)R)D)2) Reporter trust band + bookkeeping state
3) PR3 scoring behavior coverage
4) Event observability
storage_truth_score_updatedevent and attributes for:5) PR review hardening fixes included
GetParams(...).WithDefaults()store reads in inner scoring helpers by threading already-fetched params.scaleInt64TowardZero) and added edge-case coverage.Tests
Added/updated focused tests for:
Non-Goals (explicitly deferred)
Risk / Compatibility