Files
git.stella-ops.org/docs/modules/binary-index/deltasig-v2-schema.md
2026-01-20 00:45:38 +02:00

165 lines
4.1 KiB
Markdown

# DeltaSig v2 Predicate Schema
> **Sprint**: SPRINT_20260119_004_BinaryIndex_deltasig_extensions
> **Status**: Implemented
## Overview
DeltaSig v2 extends the function-level binary diff predicate with:
- **Symbol Provenance**: Links function matches to ground-truth corpus sources (debuginfod, ddeb, buildinfo, secdb)
- **IR Diff References**: CAS-stored intermediate representation diffs for detailed analysis
- **Explicit Verdicts**: Clear vulnerability status with confidence scores
- **Function Match States**: Per-function vulnerable/patched/modified/unchanged classification
## Schema
**Predicate Type URI**: `https://stella-ops.org/predicates/deltasig/v2`
### Key Fields
| Field | Type | Description |
|-------|------|-------------|
| `schemaVersion` | string | Always `"2.0.0"` |
| `subject` | object | Single subject (PURL, digest, arch) |
| `functionMatches` | array | Function-level matches with evidence |
| `verdict` | string | `vulnerable`, `patched`, `partial`, `unknown`, `partially_patched`, `inconclusive` |
| `confidence` | number | 0.0-1.0 confidence score |
| `summary` | object | Aggregate statistics |
### Function Match
```json
{
"functionId": "sha256:abc123...",
"name": "ssl_handshake",
"address": 4194304,
"size": 256,
"matchScore": 0.95,
"matchMethod": "semantic_ksg",
"matchState": "patched",
"symbolProvenance": {
"sourceId": "fedora-debuginfod",
"observationId": "obs:gt:12345",
"confidence": 0.98,
"resolvedAt": "2026-01-19T12:00:00Z"
},
"irDiff": {
"casDigest": "sha256:def456...",
"statementsAdded": 5,
"statementsRemoved": 3,
"changedInstructions": 8
}
}
```
### Summary
```json
{
"totalFunctions": 150,
"vulnerableFunctions": 0,
"patchedFunctions": 12,
"unknownFunctions": 138,
"functionsWithProvenance": 45,
"functionsWithIrDiff": 12,
"avgMatchScore": 0.85,
"minMatchScore": 0.42,
"maxMatchScore": 0.99,
"totalIrDiffSize": 1234
}
```
## Version Negotiation
Clients can request specific predicate versions:
```json
{
"preferredVersion": "2",
"requiredFeatures": ["provenance", "ir-diff"]
}
```
Response:
```json
{
"version": "2.0.0",
"predicateType": "https://stella-ops.org/predicates/deltasig/v2",
"features": ["provenance", "ir-diff"]
}
```
## VEX Integration
DeltaSig v2 predicates can be converted to VEX observations via `IDeltaSigVexBridge`:
| DeltaSig Verdict | VEX Status |
|------------------|------------|
| `patched` | `fixed` |
| `vulnerable` | `affected` |
| `partially_patched` | `under_investigation` |
| `inconclusive` | `under_investigation` |
| `unknown` | `not_affected` (conservative) |
### Evidence Blocks
VEX observations include evidence blocks:
1. **deltasig-summary**: Aggregate statistics
2. **deltasig-function-matches**: High-confidence matches with provenance
3. **deltasig-predicate-ref**: Reference to full predicate
## Implementation
### Core Services
| Interface | Implementation | Description |
|-----------|----------------|-------------|
| `IDeltaSigServiceV2` | `DeltaSigServiceV2` | V2 predicate generation |
| `ISymbolProvenanceResolver` | `GroundTruthProvenanceResolver` | Ground-truth lookup |
| `IIrDiffGenerator` | `IrDiffGenerator` | IR diff generation with CAS |
| `IDeltaSigVexBridge` | `DeltaSigVexBridge` | VEX observation generation |
### DI Registration
```csharp
services.AddDeltaSigV2();
```
Or with options:
```csharp
services.AddDeltaSigV2(
configureProvenance: opts => opts.IncludeStale = false,
configureIrDiff: opts => opts.MaxParallelism = 4
);
```
## Migration from v1
Use `DeltaSigPredicateConverter`:
```csharp
// v1 → v2
var v2 = DeltaSigPredicateConverter.ToV2(v1Predicate);
// v2 → v1
var v1 = DeltaSigPredicateConverter.ToV1(v2Predicate);
```
Notes:
- v1 → v2: Provenance and IR diff will be empty (add via resolver/generator)
- v2 → v1: Provenance and IR diff are discarded; verdict/confidence are lost
## JSON Schema
Full schema: [`docs/schemas/predicates/deltasig-v2.schema.json`](../../../schemas/predicates/deltasig-v2.schema.json)
## Related Documentation
- [Ground-Truth Corpus](./ground-truth-corpus.md)
- [Semantic Diffing](./semantic-diffing.md)
- [Architecture](./architecture.md)