rds: filter RDS_INFO_* getsockopt by caller's netns
CVE-2026-97476
rds: filter RDS_INFO_* getsockopt by caller's netns
2026-09-29 归档
今日共收录 1279 条安全情报，包含 376 个 CVE，505 篇安全通告，以及 100 篇研究论文。
← 返回归档列表RDMA/efa: Fix PBL chunk length computation
CVE-2026-93264
RDMA/efa: Fix PBL chunk length computation
smb/client: zero-initialize stack-allocated cifs_open_info_data
CVE-2026-93801
smb/client: zero-initialize stack-allocated cifs_open_info_data
media: meson: vdec: fix NULL pointer deref in vdec_try_fmt_common
CVE-2026-93236
media: meson: vdec: fix NULL pointer deref in vdec_try_fmt_common
smb: client: fix races in cifsd thread creation
CVE-2026-93829
smb: client: fix races in cifsd thread creation
wifi: rtw89: fix HE extended capability length check
CVE-2026-93281
wifi: rtw89: fix HE extended capability length check
ocfs2: fix circular locking dependency in ocfs2_init_acl()
CVE-2026-93252
ocfs2: fix circular locking dependency in ocfs2_init_acl()
regulator: tps6594: Fix device node reference leaks in multiphase loop
CVE-2026-93273
regulator: tps6594: Fix device node reference leaks in multiphase loop
signal: avoid shared siginfo namespace rewrites
CVE-2026-93222
signal: avoid shared siginfo namespace rewrites
gfs2: fix quota init duplicate scan
CVE-2026-97521
gfs2: fix quota init duplicate scan
of: fix out-of-bounds read in of_alias_scan() stem parser
CVE-2026-93213
of: fix out-of-bounds read in of_alias_scan() stem parser
CISA Says Attackers Are Exploiting Two Critical Citrix NetScaler Flaws Globally
CVE-2026-88771
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) on Sunday added two critical Citrix NetScaler ADC and G
Citrix Confirms 2 NetScaler Zero-Days After Admins Pulled the Plug
CVE-2026-88771 · CVE-2026-88772
Citrix has released patches for the critical NetScaler vulnerabilities tracked as CVE-2026-88771 and CVE-2026-88772. The
SmartMemory: Detecting On-chain-off-chain Communication Inconsistency for Smart Contract via Memory-based Agent
Smart contracts 支撑着去中心化金融 (DeFi)，而链上与链下通信 (On-chain/Off-chain Communication, OFC) 需求的增长，催生了跨链桥、真实世界资产 (RWA) 代币化、法币抵押稳定币等
What Does It Mean to Forget a Person? Individual-Level Unlearning in Vision-Language Models
这篇论文关注视觉-语言模型(VLM)中的隐私删除问题,指出现有研究与评测体系存在一个关键盲区。作者认为,从 VLM 中抹除某个具体个人的身份信息,远比抹除单条属性困难,因为个人数据并非以孤立字段形式存储,而是跨图像、文本等多种模态相互纠缠、
Physics-Attested Federated Learning: Securing Collaborative Anomaly Detection in Critical Water Infrastructure
该论文聚焦工业控制环境下的联邦学习安全问题。联邦学习允许水厂等关键基础设施运营方在不共享专有运行遥测数据的前提下，协同训练共享的入侵检测模型，从而兼顾检测能力与数据保密。但作者指出，现有防御机制几乎全部工作在“模型更新空间”，即仅对上传的梯
Leaky Students: Membership Inference against On-Policy Distillation
该论文研究的是「在线策略蒸馏（On-Policy Distillation, OPD）」场景下的成员推断攻击（Membership Inference）。OPD 的训练方式与传统蒸馏不同：它让学生模型在自己生成的轨迹上，去匹配教师模型给出的
No Free Efficiency: Revisiting the Trade-off Between Training Efficiency and Model Vulnerability
该论文研究了基础模型训练效率与模型脆弱性之间可能存在的权衡关系。近年来，为应对大规模训练带来的巨大算力与数据需求，研究者广泛采用选择性数据采样、高效预训练以及简化强化学习（RL）训练管线等策略。这些方法显著降低了开销，但作者提出一个此前被忽
AI-Based Vulnerability Assessment Capability and Cyber Attack Graph Analysis
该论文针对关键任务基础设施面临的网络威胁日益复杂、攻击门槛持续下降的现状，指出防病毒与防火墙等传统单点防护方案属于被动响应，无法应对现代攻击面的组合爆炸式复杂度。为此，作者结合两种互补的方法学开展研究：其一是洛克希德·马丁的 Vortex/
Poster: Towards ProofWeave: A Privacy-Minimised, Integrity-Anchored Evidence Plane for Continuous Agentic Assurance
该论文聚焦于智能体（Agentic AI）系统的可审计性问题。现代智能体通过工具调用、持久记忆、任务委派和外部服务来完成动作，现有的可观测性与溯源机制（日志、血缘图、追踪系统）基本只能在事后重建事件，却难以证明「在该动作执行的那一刻，是否符