Unverified50% confidenceFactExact time
鲁棒聚合方案在面对协同攻击或 Sybil 攻击时仍然脆弱，尤其当恶意方占比接近或超过 1/3 时理论安全保证将不复存在
1
Sources
50%
Confidence
Long-term
Relevance
8/19/2026
First Seen
Sources
抗投毒概念锚定：防御AI数据污染的新思路
Related Claims
Unverified在多代理架构中提示注入存在跨代理传播风险，目前尚无成熟的系统性防御方案72% similarUnverifiedThe cybersecurity domain has long faced an asymmetric offense-defense dilemma: attackers only need to find one vulnerability to succeed, while defenders must plug every vulnerability.67% similarUnverified即便出现功能强大的量子计算机，只要攻击者手中碎片数量不足 k 个，Shamir 方案依然安全，无需迁移至后量子密码方案66% similarUnverified对抗审查机制让另一个模型实例扮演攻击者专门寻找证明漏洞，只有能扛住攻击的证明才被保留66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/772864API
curl https://kongchang.com/api/v1/knowledge/claims/772864MCP
get_claim(id=772864)