Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 254papers have no tags yet and are not included in filtered results.

Attack type：Model tampering Attack type：Detection evasion Attack surface / entry point：Code repositories and dependencies Attack surface / entry point：Training data Clear all

Another 254papers do not have category tags yet and are not included in filtered results.

DefensesarXiv 2610.10345Oct 8
SLDR：通过选择性层恢复与动态路由防御恶意微调
提出 SLDR，用选择性层恢复与动态路由防御恶意微调
arXiv
2610.10345
Published
Oct 8
Layer
Training and data layer
Setting
Models and APIs
Attacker
Black-box
Evaluated models
Llama3.1-8B-Instruct, Llama3-8B-Instruct, Qwen2.5-7B-Instruct and others4
- DefensesModel hardening
- AttacksModel tampering
- ComponentFine-tuning and open weights
Abstract作者称 SLDR 能压低有害输出并保持下游效用。
SLDR 是针对恶意微调的微调后防御：在已对齐模型上做有符号层探测，只在敏感分数最大与最小的两层训练 LoRA，再用 lmax 的最后 token 表示决定是否启用适配器。
Paper analysisRead analysis
InterpretabilityarXiv 2610.07518Oct 6
TRACE：仅凭 checkpoint 更新审计大语言模型有害目标
提出 TRACE，仅凭检查点更新审计有害合规微调目标
arXiv
2610.07518
Published
Oct 6
Layer
Training and data layer
Setting
Models and APIs
Evaluated models
Llama-3.1-8B-Instruct, Mistral-7B-Instruct-v0.3, Qwen3-8B and others6
- AttacksModel tampering
- ComponentFine-tuning and open weights
Abstract作者称有害合规 SFT 在检查点更新中留下可排序坐标，并据此做只读权重审计。
作者在检查点更新里寻找有害合规 SFT 的目标组成，并把读数做成不查询模型、不看未知训练数据的审计分 TRACE。
Paper analysisRead analysis
DefensesarXiv 2609.39866Sep 30
研究者提出梯度封堵实现 LLM 发布前的预防性遗忘
提出 GSU 在发布前封堵敏感门梯度，抵抗下游微调获取禁止能力
arXiv
2609.39866
Published
Sep 30
Layer
Training and data layer
Setting
Models and APIs
Attacker
White-box
Evaluated models
Llama-3.2-1B, Llama-3.2-3B, Llama-3.1-8B and others7
- DefensesModel hardening
- AttacksModel tampering
- ComponentFine-tuning and open weights
AbstractGSU 在发布前封堵 SiLU 门的获取梯度，并在 TOFU 与 WMDP 上降低下游微调得分。
Gradient-Sealed Unlearning（GSU）是一种发布前的 preemptive unlearning：不删除模型里已经有的能力，而是让指定禁止域能力更难被发布后的微调获取。
Paper analysisRead analysis
AttacksarXiv 2609.39788Sep 30
研究：多智能体系统潜在通信链路可被训练用于提升有害顺从
提出只训练或篡改多智能体潜在通信链路来提高有害遵从
arXiv
2609.39788
Published
Sep 30
Layer
Application layer
Setting
Multi-agent systems
Evaluated models
Llama-3.2-3B-Instruct, Qwen2.5-3B-Instruct, Gemma-3-1B-IT and others7
- AttacksModel tampering
- TechniqueGradient and representation optimization
Abstract潜在链路本身即可改变冻结智能体系统的安全，奖励既能攻击也能修复。
学习得到的潜在通信链路会改变由冻结、对齐智能体组成的系统是否遵从有害请求。。作者研究三种拓扑：双智能体、三级顺序协作，以及双专家加汇总。
Paper analysisRead analysis
DefensesarXiv 2609.39279Sep 30
FDCU：用双重约束子空间投影抵御机器遗忘后的重训练攻击
提出 FDCU 双约束子空间投影遗忘，抵御再训练恢复有害知识
arXiv
2609.39279
Published
Sep 30
Layer
Training and data layer
Setting
Models and APIs
Evaluated models
Qwen2.5-3B, Llama-3-8B-Instruct, Qwen-3-8B
- DefensesModel hardening
- AttacksModel tampering
- ComponentFine-tuning and open weights
AbstractFDCU 用双掩码限制浅层抑制，在知识擦除与安全输出上提高再训练后的稳定性。
FDCU 是一种约束参数更新的遗忘方法，用来降低再训练把已抑制恶意行为重新引出的可能。
Paper analysisRead analysis
DefensesarXiv 2609.38909Sep 30
Purdue 提出 Pact：将欺骗作为行为遗忘
提出 PACT，遗忘推理模型在压力下的欺骗行为并保留上下文理解
arXiv
2609.38909
Published
Sep 30
Layer
Training and data layer
Setting
Models and APIs
Attacker
White-box
Evaluated models
DeepSeek-R1-Distill-Qwen-32B, QwQ-32B
- DefensesModel hardening
- AttacksModel tampering
- RiskDeception and scheming
- ComponentFine-tuning and open weights
Abstract论文将大模型欺骗形式化为上下文条件行为，提出 PACT 框架在消除欺骗的同时保留了上下文遵循能力。
Paper analysisRead analysis
DefensesarXiv 2609.36862Sep 29
VaccineBooster：面向有害微调对齐的混合扰动防御
提出 VaccineBooster，在对齐阶段混合嵌入扰动与梯度衰减抵御有害微调
arXiv
2609.36862
Published
Sep 29
Layer
Training and data layer
Setting
Models and APIs
Evaluated models
Llama-2-7B
- DefensesModel hardening
- AttacksModel tampering
- ComponentFine-tuning and open weights
AbstractVaccineBooster 把两种对齐阶段扰动合在一步里，内容更安全但拒答保留更少。
VaccineBooster 是一种对齐阶段防御：在提供方还看得到训练过程、还没接触用户微调数据时，把嵌入扰动和权重级梯度衰减合成同一次更新，用来应对之后不受信任的微调。有害微调可以只混入少量有害指令-回答对，就削弱拒答并提高有害输出的可能，任务效用却仍然好看。过滤会漏掉隐蔽样本，事后修复又往往不知道该修哪里。
Paper analysisRead analysis
DefensesarXiv 2610.00320Sep 29
研究显示少样本微调可绕过安全层局部冻结与修复防御
检验层冻结与奇异方向截断能否挡住少样本有害微调
arXiv
2610.00320
Published
Sep 29
Layer
Training and data layer
Setting
Models and APIs
Attacker
White-box
Evaluated models
Llama-3.1-8B-Instruct, Llama-3.1-Tulu-3-8B-DPO, OLMo-2-1124-7B-Instruct and others10
- DefensesModel hardening
- AttacksModel tampering
- ComponentFine-tuning and open weights
Abstract定位与恢复可以复现，但知情攻击能把损伤移出冻结带，并打败跨检查点的截断修复。
这篇工作检验：把拒答定位到某些层或更新方向之后，能不能在攻击者知情时还守住拒答。。
Paper analysisRead analysis
AttacksarXiv 2609.27124Sep 23
Fed-ADR：用恶意编排服务器协同客户端梯度操纵攻击联邦学习
提出 Fed-ADR，用编排服务器协同操纵梯度攻破联邦学习并检测恢复
arXiv
2609.27124
Published
Sep 23
Layer
Training and data layer
Setting
Models and APIs
Attacker
Gray-box
Evaluated models
CNN（MNIST：2 Conv + 3 FC）, CNN（Fashion-MNIST：6 Conv + 1 FC）, ResNet18（CIFAR-10）
- DefensesDetection and filtering
- AttacksPoisoning and backdoors
- TechniqueDefense-targeted
AbstractFed-ADR 用 gray-box 编排攻击绕过多种聚合防御，再用 Hessian 一致性检测和窗口内恢复把准确率拉回。
Fed-ADR 是一套面向联邦学习的攻击、检测与恢复框架：编排服务器改写合谋客户端的梯度，使参数服务器的现有防御把投毒更新当成良性，同时在发现后用历史梯度把全局模型原地修回来。
Paper analysisRead analysis
InterpretabilityarXiv 2609.06934Sep 7
论文提出拒答的几何解释：事后安全训练脆弱，预训练期持续安全共训更抗攻击
解释后验安全的权重几何脆弱性，并提出预训练持续安全共训练
arXiv
2609.06934
Published
Sep 7
Layer
Model layer
Setting
Models and APIs
Evaluated models
Qwen-2.5-7B, Qwen-2.5-7B-Instruct, Llama-3-8B and others15
- DefensesModel hardening
- AttacksModel tampering
- RiskInappropriate refusal
- ComponentFine-tuning and open weights
Abstract论文将后验安全脆弱性归因于正交抑制，并提出通过预训练持续共训练实现具备微调抗性的安全防护。
Paper analysisRead analysis
DefensesarXiv 2609.02293Sep 2
SEAL：通过共享专家对齐强化 MoE 全局安全
提出 SEAL，在 MoE 共享专家上做 DPO-LoRA 对齐以抵御越狱与权重篡改
arXiv
2609.02293
Published
Sep 2
Layer
Training and data layer
Setting
Models and APIs
Attacker
White-box, Black-box
Evaluated models
Qwen1.5-MoE-A2.7B, DeepSeekMoE-16B, GLM-4.7-Flash and others4
- DefensesModel hardening
- AttacksModel tampering
AbstractSEAL 把 DPO-LoRA 限制在混合 MoE 共享专家上，SEAL++再保护已有安全神经元方向。
SEAL是面向混合 MoE 的训练时防御：只对齐始终执行的共享专家，使安全不依赖路由器是否把 token 送到安全专家。
Paper analysisRead analysis
DefensesarXiv 2608.06422Aug 6
CMU 研究：分片判断可避免 LLM 监督失效并抵御对抗利用
提出分片监督，将评判标准分配给独立调用以抵御展示攻击
arXiv
2608.06422
Published
Aug 6
Layer
Application layer
Setting
LLM applications
Attacker
Black-box
Evaluated models
Claude Opus 4.8, Claude Sonnet 4.6, Claude Opus 4.6 and others4
- DefensesDetection and filtering
- AttacksDetection evasion
- ComponentGuardrails and judges
Abstract论文揭示决策负载是大模型监督的独立瓶颈，提出分片与辩论机制有效分离注意力与平衡证据，在多项高难度基准上实现稳健评判。
本文系统研究了大语言模型在执行多标准监督与评判时的决策负载瓶颈及其安全风险。
Paper analysisRead analysis
DefensesarXiv 2608.05045Aug 6
Gradient Immunity 提出 USG：在开放权重发布时阻断恶意微调梯度
提出 USG，在部分保护开源权重发布时阻断有害微调梯度
arXiv
2608.05045
Published
Aug 6
Layer
Training and data layer
Setting
Models and APIs
Attacker
White-box
Evaluated models
Qwen3-14B（表中作 Qwen-14B）, Llama-3.1-8B（表中作 Llama-8B）
- DefensesModel hardening
- AttacksModel tampering
- ComponentFine-tuning and open weights
AbstractUSG 在 PPOW 下用零空间立方层阻断有害微调梯度。
Unidirectional Safety Gate 是面向 PPOW 发布设定的发布时防护：大部分权重仍可训练，一小部分安全关键组件由提供者侧加固保留，下游不必配合额外安全流程。。
Paper analysisRead analysis
AttacksarXiv 2608.03642Aug 4
规避与投毒攻击对恶意软件数据漂移检测器的实证分析
用通用扰动对恶意软件分类器及漂移检测器做规避与后门投毒
arXiv
2608.03642
Published
Aug 4
Layer
Model layer
Setting
Models and APIs
Evaluated models
MLP 100-100, CADE (CAE 512-384-256-128), CADE (CAE 512-128-32-8) and others6
- AttacksDetection evasion
- TechniqueGradient and representation optimization
Paper analysisRead analysis
AttacksarXiv 2605.30040May 28
研究揭示 LLM 按 token 计费可被服务商系统性虚报
构造可通过计费审计的 token 虚报，使服务商系统性超收
arXiv
2605.30040
Published
May 28
Layer
Infrastructure layer
Setting
Models and APIs
Evaluated models
PALACE domain-adapted auditor
- AttacksDetection evasion
- TechniqueDefense-targeted
- ComponentMonitors
Abstract不诚实提供方可在三种 token 审计上虚报并通过检查，作者认为需改用提供方不控制的证据。
不透明 LLM 的按 token 计费里，现有审计核验的是提供方可控的证据，不诚实提供方可以虚报并通过检查。。
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2605.14605May 14
Study: malicious fine-tuning defenses fail under continued training, harmfulness rises across all 72 defense trajectories
用代价曲线评估恶意微调防御在持续训练下的衰减
arXiv
2605.14605
Published
May 14
Layer
Training and data layer
Setting
Models and APIs
Evaluated models
Llama-2-7B-chat, Qwen3-8B, Llama-3.1-8B-Instruct and others8
- DefensesModel hardening
- AttacksModel tampering
- ComponentFine-tuning and open weights
Abstract论文指出恶意微调防御在持续训练下均会衰减，表明基于有限攻击者假设的防御无法为开源模型提供持久保护。
Paper analysisRead analysis
Analysis and theoryarXiv 2604.02574Apr 3
Understanding the impact of safety unalignment on large language models
Comparing jailbreak fine-tuning and weight orthogonalization unalignment on harmful assistance and usefulness
arXiv
2604.02574
Published
Apr 3
Layer
Model layer
Setting
Models and APIs
Evaluated models
Qwen3-4B-Instruct-2507, Llama-3.1-8B-Instruct, Qwen2.5-14B and others6
- AttacksModel tampering
- ComponentFine-tuning and open weights
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2510.14207Oct 16, 2025
论文提出多轮在线骚扰 Agent 基准
提出多轮在线骚扰 Agent 基准，评测记忆、规划与微调越狱
arXiv
2510.14207
Published
Oct 16, 2025
Layer
Application layer
Setting
Multi-agent systems
Evaluated models
LLaMA-3.1-8B-Instruct, Qwen3-30B-A3B-Instruct, DeepSeekMoE-16B and others4
- AttacksJailbreaks
- ComponentFine-tuning and open weights
Paper analysisRead analysis
AttacksarXiv 2510.02833Oct 3, 2025
研究者提出用 10 条良性问答对微调越狱 LLM 的两阶段攻击
提出两阶段良性微调攻击，经 FaaS 用良性数据解除安全对齐
arXiv
2510.02833
Published
Oct 3, 2025
Layer
Training and data layer
Setting
Models and APIs
Evaluated models
Llama2-7b-chat-hf, Llama3-8b-instruct, Deepseek-R1-Distill-Llama3-8b and others10
- AttacksModel tampering
- TechniqueDefense-targeted
- ComponentFine-tuning and open weights
Paper analysisRead analysis

You have reached the end

AI Safety Daily

Read the latest issueOne issue per day, updated as events develop