Files
nexus/wiki/concepts/Trust-Scoring.md

55 lines
1.7 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
---
title: "Trust-Scoring"
type: concept
tags: [trust, scoring, agent-reliability]
sources: [agentic-identity-trust.md]
last_updated: 2026-04-25
---
## Definition
Trust-Scoring 是一种基于**可验证结果**的惩罚型信任量化模型。Agent 从初始分数 1.0 开始仅通过可客观验证的问题进行扣分Agent 不得自我报告信号来提高信任分。
## Scoring Model
```python
score = 1.0
# 证据链损坏(最高惩罚)
if not check_chain_integrity(agent_id):
score -= 0.5
# 结果验证(失败率 × 0.4
outcomes = get_verified_outcomes(agent_id)
if outcomes.total > 0:
failure_rate = 1.0 - (outcomes.achieved / outcomes.total)
score -= failure_rate * 0.4
# 凭证新鲜度
if credential_age_days(agent_id) > 90:
score -= 0.1
```
## Trust Levels
| 分数区间 | 信任等级 | 说明 |
|---------|---------|------|
| ≥ 0.9 | HIGH | 完全可信任,可执行高风险操作 |
| ≥ 0.5 | MODERATE | 有限信任,需额外验证 |
| > 0.0 | LOW | 高度谨慎,需全面验证 |
| 0.0 | NONE | 不可信,拒绝所有操作请求 |
## Key Properties
- **无自我报告**Agent 不能声称自己是可信的——信任来自客观可验证的证据
- **信任衰减**:长期未活动的 Agent 或过期凭证自动降低信任分
- **证据驱动**:评分基于 [[Evidence-Chain]] 中的实际执行结果,而非声誉或声明
## Relationships
- [[Zero-Trust]]Trust-Scoring 是 Zero-Trust 可量化验证的实现
- [[Evidence-Chain]]Trust-Scoring 的评分依据来源
- [[Peer-Verification]]Peer-Verification 协议使用 Trust-Scoring 作为决策依据之一
## Sources
- [[agentic-identity-trust.md]]