chg_0011(T1): docs/dth/vc/scenario.md — 6 步流程 + 6 算子映射 + 出域 PII 清单
This commit is contained in:
1 parent
55e93b25f0
commit
48589b3c1d
1 file changed
+366
@@ -0,0 +1,366 @@
|
||||
# Vehicle-Cloud Data Compliance (VC) — 车云数据合规场景设计
|
||||
|
||||
> chg_0011 · 2026-09-29 · owner: pm@team · 状态:implemented
|
||||
>
|
||||
> **0 新代码 / 0 新依赖 / 0 新接口**。本场景把 chg_0008 / chg_0009 已落地的
|
||||
> 6 个算子按 6 步流程组合起来,证明 SP-1 实验台可在真实"数据出域可追溯"流程下
|
||||
> 协同工作,满足"出域 PII 必脱敏 + 可还原 + 可审计"的车云合规要求。
|
||||
>
|
||||
> **与 chg_0010 的关系**:chg_0010 = 数据不出域(金融联合风控),
|
||||
> chg_0011 = **数据出域可追溯**(车云合规)。两类场景互补覆盖需求 §5 的
|
||||
> "业务场景测试"全部形态。
|
||||
|
||||
## 1. 业务背景
|
||||
|
||||
智能网联车的运行数据需要离开车端到云端做合规分析(监管 / 远程诊断 / OTA 监控),
|
||||
但**任何带 PII 的字段必须脱敏 + 审计可还原**,且全程要留不可篡改的证据链。
|
||||
|
||||
50 辆车(vehicle-001 ~ vehicle-050)每辆都持有:
|
||||
|
||||
| 字段类别 | 字段名 | 是否出域 | 出域形态 |
|
||||
|---|---|---|---|
|
||||
| PII | GPS 经纬度 | ❌ | 脱敏为 vehicle-region 标签 |
|
||||
| PII | 车牌号 | ❌ | 脱敏保留前 2 后 1(如"京A****1")|
|
||||
| PII | 车主 ID(身份证)| ❌ | 脱敏保留前 6 后 4 |
|
||||
| 业务 | 故障码(如 P0301)| ✅ | 明文上云 |
|
||||
| 业务 | 电池电压(12.6V)| ✅ | 明文上云 |
|
||||
| 业务 | 累计里程(km)| ✅ | 明文上云 |
|
||||
|
||||
业务目标:50 辆车在不暴露 PII 明文的前提下,完成一次端云协同合规分析
|
||||
(**`TDS 合约 → 字段脱敏 → 报告水印 → 车队聚合 → ABAC 鉴权 → 审计链`**),
|
||||
并全程留下可还原 + 不可篡改的证据。
|
||||
|
||||
## 2. mermaid 流程图(6 步串联)
|
||||
|
||||
```mermaid
|
||||
flowchart TD
|
||||
subgraph S1[Step 1 — 车端 TDS 合约]
|
||||
V[vehicle-001 ~ vehicle-050]
|
||||
TDS[tds-connector.match<br/>contract_hash 双方一致<br/>→ JWT-like token]
|
||||
V --> TDS
|
||||
end
|
||||
|
||||
subgraph S2[Step 2 — 拒绝路径]
|
||||
TDM[tds-connector.match<br/>contract_hash_a ≠ contract_hash_b<br/>→ error=contract-mismatch]
|
||||
TDM -. 拒绝 .-> ABORT[流程中止<br/>无后续步骤]
|
||||
end
|
||||
S1 --> S2
|
||||
|
||||
subgraph S3[Step 3 — 字段感知脱敏]
|
||||
MASK_GPS[data-masking g 位位置]
|
||||
MASK_PLATE[data-masking 车牌]
|
||||
MASK_OWNER[data-masking 车主ID]
|
||||
BIZ[业务字段 P0301 / 12.6V / 50000km<br/>明文上云]
|
||||
MASK_GPS & MASK_PLATE & MASK_OWNER --> HASH[reversible_hash<br/>SHA256 每字段]
|
||||
end
|
||||
S2 --> S3
|
||||
|
||||
subgraph S4[Step 4 — 报告水印]
|
||||
WM[data-watermark<br/>payload=VC|2026-09-29|fleet-50|v1<br/>→ 零宽水印嵌入]
|
||||
EX[data-watermark.extract<br/>→ 还原 payload]
|
||||
WM --> EX
|
||||
end
|
||||
S3 --> S4
|
||||
|
||||
subgraph S5[Step 5 — 车队聚合]
|
||||
FV[horizontal-fedavg<br/>50 辆车 故障率均值<br/>→ 车队级 fault_rate]
|
||||
end
|
||||
S4 --> S5
|
||||
|
||||
subgraph S6[Step 6 — ABAC 鉴权 + 审计链]
|
||||
ABAC[abac-policy<br/>role=compliance-analyst<br/>ip in 10.0.0.0/24]
|
||||
AUD[compliance-audit<br/>6 步流程 HMAC chain<br/>→ chainValid=true]
|
||||
ABAC --> AUD
|
||||
end
|
||||
S5 --> S6
|
||||
```
|
||||
|
||||
## 3. 数据流表(每步:输入 → 算子 → 输出 → 不变性)
|
||||
|
||||
| Step | 输入 | 算子 | 输出 | 不变性 |
|
||||
|---|---|---|---|---|
|
||||
| S1 | `contract_hash_a = "vc-contract-v1"`, `contract_hash_b = "vc-contract-v1"` | tds-connector | `{contract_hash_match: true, access_token: <3-part JWT>}` | JWT 三位段(header.payload.signature) |
|
||||
| S2 | `contract_hash_a = "vc-contract-v1"`, `contract_hash_b = "vc-contract-v2"` | tds-connector | `{contract_hash_match: false, error: "contract-mismatch", access_token: null}` | 无 token = 流程中止 |
|
||||
| S3 | `{field_type: "gps", value: "<lat,lon>"}` × 50 | data-masking | `{masked: "vehicle-region-NN", reversible_hash: <sha256-hex>}` × 50 | 无原始 PII;SHA256 64 hex |
|
||||
| S4 | `text: "fleet report..."`, `payload: "VC|2026-09-29|fleet-50|v1"` | data-watermark (embed + extract) | `{watermarked_text, extracted_payload}` | roundtrip payload 字节一致 |
|
||||
| S5 | `weights: [0.02, 0.05, ..., 0.08]`(50 个)| horizontal-fedavg | `{aggregated: [mean], client_count: 50}` | 无 epsilon → 精确均值 |
|
||||
| S6 | `{subject: {role}, env: {ip}}, policy: {...}` | abac-policy | `{decision: allow/deny, reasons: [...]}` | allow iff role+ip 都匹配 |
|
||||
| S6 | `{action, actor, chain: [...prior entries]}` × 6 | compliance-audit | `{entry: {prev_hash, hash}} × 6`, `chainValid: true` | HMAC 链 |
|
||||
|
||||
## 4. 算子映射表
|
||||
|
||||
### 4.1 tds-connector(S1 + S2)
|
||||
|
||||
- **何时调用**:车端 TDS 客户端 → 云端合约;云端 `contract_hash` 与车端 `contract_hash` 比对。
|
||||
- **input schema**:
|
||||
```json
|
||||
{
|
||||
"contract_hash_a": "string", // 必填
|
||||
"contract_hash_b": "string", // 必填
|
||||
"subject": "string", // JWT sub(默认 "party-unknown")
|
||||
"scope": "string", // 数据访问 scope(默认 "read")
|
||||
"ttl_seconds": 3600 // token TTL(默认 3600)
|
||||
}
|
||||
```
|
||||
- **output schema(match=true)**:
|
||||
```json
|
||||
{
|
||||
"ok": true,
|
||||
"tds": {
|
||||
"contract_hash_match": true,
|
||||
"sha256_a": "<hex>",
|
||||
"sha256_b": "<hex>",
|
||||
"access_token": "<header>.<payload>.<sig>",
|
||||
"token_type": "JWT-like",
|
||||
"ttl_seconds": 3600,
|
||||
"subject": "vehicle-cloud-001",
|
||||
"scope": "read",
|
||||
"algorithm": "tds-hmac-jwt-v2"
|
||||
},
|
||||
"schema_version": "v2"
|
||||
}
|
||||
```
|
||||
- **output schema(match=false)**:
|
||||
```json
|
||||
{
|
||||
"ok": false,
|
||||
"tds": {
|
||||
"contract_hash_match": false,
|
||||
"error": "contract-mismatch",
|
||||
"access_token": null
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
### 4.2 data-masking(S3 + S4)
|
||||
|
||||
- **何时调用**:车端上报 PII 字段(GPS / 车牌 / 车主 ID),每字段一次调用。
|
||||
- **input schema**:
|
||||
```json
|
||||
{
|
||||
"field_type": "phone|id_card|email|bank_card|name",
|
||||
"value": "string"
|
||||
}
|
||||
```
|
||||
- **output schema**:
|
||||
```json
|
||||
{
|
||||
"ok": true,
|
||||
"masking": {
|
||||
"field_type": "id_card",
|
||||
"original_length": 18,
|
||||
"masked": "110101********1234",
|
||||
"masked_length": 18,
|
||||
"reversible_hash": "<64-hex sha256>",
|
||||
"algorithm": "field-aware-masking-v2"
|
||||
}
|
||||
}
|
||||
```
|
||||
- **VC 用法**:VC 把 5 个内置类型扩展为 3 个车云 PII 类型(**gps / plate / owner_id**),
|
||||
复用**通用"保留头部+尾部"规则**,并对**业务字段跳过调用**(直接明文上云)。
|
||||
|
||||
### 4.3 data-watermark(S5 + S6)
|
||||
|
||||
- **何时调用**:fleet 聚合报告生成后,在报告 body 嵌入零宽水印。
|
||||
- **input schema(embed)**:
|
||||
```json
|
||||
{ "action": "embed", "text": "string", "payload": "string" }
|
||||
```
|
||||
- **input schema(extract)**:
|
||||
```json
|
||||
{ "action": "extract", "text": "<watermarked_text>" }
|
||||
```
|
||||
- **output schema(embed)**:
|
||||
```json
|
||||
{
|
||||
"watermark": {
|
||||
"action": "embed",
|
||||
"text": "...",
|
||||
"payload": "...",
|
||||
"watermarked_text": "<text + 4*payload.length 零宽字符>",
|
||||
"zero_width_chars_appended": 96, // = 24 * 4
|
||||
"length_invariant_ok": true
|
||||
}
|
||||
}
|
||||
```
|
||||
- **output schema(extract)**:
|
||||
```json
|
||||
{ "watermark": { "action": "extract", "extracted_payload": "..." } }
|
||||
```
|
||||
|
||||
### 4.4 horizontal-fedavg(S7)
|
||||
|
||||
- **何时调用**:50 辆车的故障率数据收集完毕后,调用一次聚合。
|
||||
- **input schema**(v2 新模式 — chg_0008 S2):
|
||||
```json
|
||||
{ "weights": [0.02, 0.05, ..., 0.08] } // 50 个 [0,1] 数值
|
||||
```
|
||||
- **output schema**:
|
||||
```json
|
||||
{
|
||||
"ok": true,
|
||||
"fedavg": {
|
||||
"aggregated": [<mean>],
|
||||
"final_weights": <mean>,
|
||||
"dimension": 1,
|
||||
"client_count": 50,
|
||||
"round_count": 1,
|
||||
"aggregation": "mean",
|
||||
"algorithm": "fedavg-gaussian-dp-v2",
|
||||
"noise_scale": 0, // 无 epsilon
|
||||
"noise_applied": false
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
### 4.5 abac-policy(S8)
|
||||
|
||||
- **何时调用**:合规分析师登录 → 请求访问车云分析报告时鉴权。
|
||||
- **input schema**:
|
||||
```json
|
||||
{
|
||||
"subject": { "role": "compliance-analyst", "name": "alice" },
|
||||
"resource": { "type": "vc-report" },
|
||||
"env": { "ip": "10.0.0.5", "time": "10:00" },
|
||||
"policy": {
|
||||
"allow_if": { "subject.role": "compliance-analyst" },
|
||||
"env_constraints": { "ip_range": "10.0.0.0/24" }
|
||||
}
|
||||
}
|
||||
```
|
||||
- **output schema**:
|
||||
```json
|
||||
{
|
||||
"abac": {
|
||||
"decision": "allow|deny",
|
||||
"reasons": ["allow_miss:..."], // 仅 deny 时填
|
||||
"allow_matched_keys": [...],
|
||||
"deny_matched_keys": [...],
|
||||
"env_violations": [...]
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
### 4.6 compliance-audit(S9)
|
||||
|
||||
- **何时调用**:6 步流程每步都 append 一条 entry;最后 verify 整链。
|
||||
- **input schema(record / append)**:
|
||||
```json
|
||||
{
|
||||
"action": "tds-match|masking|watermark-embed|fedavg|abac-allow|abac-deny",
|
||||
"actor": "vc-orchestrator",
|
||||
"user": "vehicle-001",
|
||||
"entries": [...prior entries, optional] // 续链用
|
||||
}
|
||||
```
|
||||
- **input schema(verify)**:
|
||||
```json
|
||||
{
|
||||
"verify": true,
|
||||
"entries": [<full chain from genesis>]
|
||||
}
|
||||
```
|
||||
- **output schema(record)**:
|
||||
```json
|
||||
{
|
||||
"audit": {
|
||||
"trail_id": "<uuid>",
|
||||
"entry": {
|
||||
"index": 3,
|
||||
"prev_hash": "<sha256>",
|
||||
"payload": {...},
|
||||
"hash": "<sha256>"
|
||||
},
|
||||
"chain_index": 3,
|
||||
"mode": "append" // 或 "genesis"
|
||||
}
|
||||
}
|
||||
```
|
||||
- **output schema(verify)**:
|
||||
```json
|
||||
{ "audit": { "chainValid": true, "entries_verified": 6, "first_bad_index": -1, "mode": "verify" } }
|
||||
```
|
||||
|
||||
## 5. 不变性清单(S1-S10 → 对应算子 + 测试断言)
|
||||
|
||||
| Scenario | 不变性 | 算子 | 测试断言(C 编号) |
|
||||
|---|---|---|---|
|
||||
| S1 | TDS 合约匹配 → 3-part JWT | tds-connector | C1: token 含 2 个 `.` |
|
||||
| S2 | TDS 合约不匹配 → 拒绝 | tds-connector | C2: error="contract-mismatch" |
|
||||
| S3 | 50 车 × 3 字段感知脱敏 | data-masking | C3: 无 PII 明文 + 业务字段保留 |
|
||||
| S4 | 脱敏可逆性(SHA256) | data-masking | C4: reversible_hash 64 hex + 幂等 |
|
||||
| S5 | 水印嵌入 + 提取 roundtrip | data-watermark | C5: extract === payload |
|
||||
| S6 | 水印不可见 | data-watermark | C6: visible length 不变 |
|
||||
| S7 | 50 车 fleet 均值 | horizontal-fedavg | C7: mean 精确 + client_count=50 |
|
||||
| S8 | ABAC 合规分析师通过 + 工程师拒绝 | abac-policy | C8: 1 allow + 2 deny |
|
||||
| S9 | 审计链 HMAC chainValid | compliance-audit | C9: chain length=6 + chainValid=true |
|
||||
| S10 | E2E 6 步全程通过 | 全部 6 算子 | C10: pass=10 fail=0 + < 60s |
|
||||
|
||||
## 6. 出域 PII 清单(合规关键)
|
||||
|
||||
这是本场景**与 chg_0010 的核心差异** — chg_0010 守护"不出域",本 chg_守护"**出域可追溯**"。
|
||||
|
||||
| 字段 | 是否出域 | 出域形态 | 脱敏方式 | 反向验证手段 |
|
||||
|---|---|---|---|---|
|
||||
| **GPS 经纬度** | ❌(脱敏后出域)| `vehicle-region-NN`(区域标签)| data-masking 通用规则:保留前 N 位 + mask 尾部 | reversible_hash (SHA256 字段级) → 业务对账时验证"原始 GPS hash 等于此值" |
|
||||
| **车牌号** | ❌(脱敏后出域)| `京A****1`(保留前 2 后 1)| data-masking 通用规则:保留头尾 + mask 中间 | 同上 reversible_hash |
|
||||
| **车主身份证号** | ❌(脱敏后出域)| `110101********1234`(保留前 6 后 4)| data-masking 通用规则:保留头尾 + mask 中间 | 同上 reversible_hash |
|
||||
| **故障码(P0301)**| ✅ 明文上云 | `P0301` | 不脱敏(无 PII)| N/A |
|
||||
| **电池电压(12.6V)**| ✅ 明文上云 | `12.6V` | 不脱敏(无 PII)| N/A |
|
||||
| **累计里程(50000km)**| ✅ 明文上云 | `50000km` | 不脱敏(无 PII)| N/A |
|
||||
|
||||
**核心合规承诺**:
|
||||
|
||||
1. 原始 PII(GPS / 车牌 / 车主 ID)**从不**进入云端存储;只输出**脱敏值 + SHA256 哈希**。
|
||||
2. 业务方对账时,可通过 `reversible_hash` 验证"原始 PII 是否为此值"(无需保留 PII 明文)。
|
||||
3. 整个报告作 zero-width 字符水印(嵌入 `VC|2026-09-29|fleet-50|v1`),任何云端泄露
|
||||
都能追溯到出域时间和车队批次。
|
||||
4. 6 步流程的每步都 append 一条 HMAC 审计 entry,任何流程篡改都会被 `chainValid=false` 检出。
|
||||
|
||||
## 7. 复用矩阵
|
||||
|
||||
| 算子 | chg 来源 | 复用步骤 |
|
||||
|---|---|---|
|
||||
| **tds-connector** | chg_0009 真算法 | S1 + S2(chg_0010 也用)|
|
||||
| **data-masking** | chg_0009 真算法 | S3 + S4(**chg_0011 独占**)|
|
||||
| **data-watermark** | chg_0009 真算法 | S5 + S6(chg_0010 也用)|
|
||||
| **horizontal-fedavg** | chg_0008 真算法 | S7(**chg_0011 独占**)|
|
||||
| **abac-policy** | chg_0009 真算法 | S8(chg_0010 也用)|
|
||||
| **compliance-audit** | chg_0008 真算法 | S9(chg_0010 也用)|
|
||||
|
||||
- **本 chg 用 6 个算子** = 19 总算子的 **32%**。
|
||||
- **与 chg_0010 共用 4 个**(tds / watermark / abac / audit)。
|
||||
- **本 chg 独占 2 个**(data-masking / horizontal-fedavg)。
|
||||
- **chg_0010 独占 4 个**(psi / paillier-he / threshold-sig / did-identity)。
|
||||
|
||||
## 8. 与 chg_0010 的互补关系
|
||||
|
||||
| 维度 | chg_0010 金融联合风控 | chg_0011 车云数据合规 |
|
||||
|---|---|---|
|
||||
| 数据流向 | 不出域(多方 PSI / 同态 / 门限)| **出域可追溯**(脱敏 + 水印 + 审计)|
|
||||
| 隐私保护核心 | 密码学(PSI / Paillier / Threshold Sig)| **运维/合规**(masking + watermark + audit chain)|
|
||||
| 算子覆盖率 | 8/19 = 42% | **6/19 = 32%** |
|
||||
| 业务形态 | 银行 → 银行(B2B 多方)| **车 → 云**(B2C 单方 + 多车)|
|
||||
| 失败模式 | 数据泄露 → 隐私事故 | 数据泄露 → **追溯源头 + 合规追责** |
|
||||
| 监管语境 | GDPR / 数据最小化 | **个保法 / 车联网数据安全** |
|
||||
|
||||
**两类场景合起来 = "数据流通 + 数据保护" 的完整闭环**:
|
||||
- 出域前:**chg_0010**(不让数据出门);
|
||||
- 出域时:**chg_0011**(出门必须可追溯 + 可还原 + 可鉴权)。
|
||||
|
||||
## 9. 风险与缓解
|
||||
|
||||
- **R1**: data-masking v2 是**字段级单值**算子,50 车 telemetry 是批量数据。
|
||||
- 缓解:shell 循环 50 × 3 = 150 次算子调用,保持业务流真实(每车 3 字段)。
|
||||
- **R2**: horizontal-fedavg chg_0008 测试断言 `aggregated == [0.3, 0.4]` 精确等于数学均值。
|
||||
- 缓解:VC 不传 epsilon → 走精确均值路径,断言 `aggregated[0] == mean(50 weights)`。
|
||||
- **R3**: compliance-audit verify 链需要 6 个完整 entry。
|
||||
- 缓解:S9 子步骤分 5 次 append + 1 次 verify,每步独立记 assert。
|
||||
|
||||
## 10. 总结
|
||||
|
||||
- **6 步流程** = tds → masking → watermark → fedavg → abac → audit。
|
||||
- **6 算子复用** = 19 总算子的 **32%**。
|
||||
- **10 个 e2e case** 覆盖 S1-S10,含 4 个不变性断言 + 1 个拒绝路径 + 1 个跨车 E2E。
|
||||
- **0 新代码 / 0 新依赖 / 0 新接口**。
|
||||
- **与 chg_0010 互补**:出域前(chg_0010)+ 出域时(chg_0011)= 数据流通完整闭环。
|
||||
Reference in new issue
Block a user