Instruction Alignment for Binary Code Representation Learning
二进制代码表示学习是软件安全与逆向工程中的基础问题。现有主流方法主要学习函数级别的嵌入表示，以捕捉二进制函数之间粗粒度的语义关系，但往往忽略了更细粒度的指令级对应关系。这一局限导致模型无法充分利用编译器调试信息中蕴含的丰富监督信号，而这些信
2026-08-13 归档
今日共收录 1881 条安全情报，包含 959 个 CVE，493 篇安全通告，以及 113 篇研究论文。
← 返回归档列表Defending against Model Extraction for GNNs with Model Reprogramming
本文提出一种针对图神经网络（GNN）模型窃取攻击的主动防御框架 GraphRP。在机器学习即服务（MLaaS）场景中，GNN 作为高价值模型常以黑盒 API 形式部署，攻击者可通过查询接口复制模型功能，从而窃取知识产权。现有防御手段多从图像
Association-based Privacy Attacks in Wireless Protocols: Formal Modeling and Mitigation
本文针对无线通信协议中的配对隐私威胁进行了形式化建模与缓解研究。随着社交媒体和物联网设备产生的隐私敏感数据激增，现有系统缺乏自动化、形式化的隐私风险评估方法。作者聚焦于利用重放/中继技术进行的配对关联攻击，分析了条件无关响应、重放抵抗和距离
Battlefield 5G: Dual-PKI and TPM-Based UE Attestation for Tactical 5G Standalone Networks
该论文针对战术5G独立组网（SA）环境中标准5G-AKA认证机制的不足，提出了一种名为Battlefield 5G的预认证框架。标准5G-AKA仅验证USIM中的订阅者凭据，而不验证持有该凭据的物理设备身份及其启动状态，这在设备可能被捕获、
Dual-Adversarial Safety Alignment: Cultivating Intrinsic Threat Comprehension in LRMs
本文针对大型推理模型（LRMs）在复杂任务上表现卓越但易受有害提示诱导、产生不安全输出的问题展开研究。现有对齐方法通常依赖直接拒绝或安全推理，但往往聚焦于提示模式而非内在攻击机制，导致在面对多样化越狱攻击时泛化能力不足，难以兼顾对抗鲁棒性与
Joint Lyapunov Certificates for K-Agent Generative AI Governance: Stochastic Stability, Emergent Ensemble Risk, and Zero-Knowledge Governance Attestation
本文提出一个面向多智能体生成式AI系统治理的严格数学框架，聚焦于K个自适应性生成式AI模型在模型风险管理（MRM）原则下的联合稳定性与零知识证明问题。研究背景在于：当多个模型通过交互矩阵形成元学习耦合时，传统基于单个智能体的Lyapunov
A Complexity-Theoretic Approach to Proofs of Space
本文提出了一种基于复杂性理论构建空间证明（Proof of Space, PoS）的新框架。空间证明由Dziembowski等人于CRYPTO 2015引入，是一个两阶段协议，允许证明者向高效的验证者证明其已分配大量持久性内存来存储某些信息
IoXT: The Internet of Explainable Things. Why Explainability in IoT Requires a New System-Level Paradigm and Protocol Design
本文提出“可解释物联网”（IoXT, Internet of Explainable Things）这一系统级范式，旨在使可解释性成为智能物联网的内建架构属性，而非事后附加能力。传统可解释人工智能（XAI）方法只解释模型预测，无法覆盖从传感
Strategies to Avoid Illegal Data Access
本文献围绕数据安全展开，指出防止未授权数据访问是各类规模企业面临的核心挑战。随着技术持续演进，数据被非法访问的风险不断上升，因此企业需采取主动防御策略。研究系统梳理了三大类防护措施：技术解决方案、人员培训与政策执行。技术层面重点讨论防火墙、
Track me if you can: Ephemeral coin tracing
隐私保护支付系统在理论上已相当成熟，但在中心化数字货币（CBDC）、机构稳定币及其他合规支付基础设施等受监管场景中的采用，长期受制于其对非法活动可能被滥用的担忧。监管机构通常依赖识别、追踪和阻止犯罪者的一系列互补措施来打击金融犯罪，其中“追
IBM DataPower Gateway 11.0.0.0 through 11.0.0.1 and IBM DataPower Gateway 10.5.0.0 through 10.5.0.21 and IBM DataPower Gateway 10.6.0.0 through 10.6.0.9 allows a race condition ...
CVE-2026-7366
CVE-2026-7366 是 IBM DataPower Gateway 中存在的一个竞态条件（race condition）漏洞。受影响版本包括 11.0.0.0 至 11.0.0.1、10.5.0.0 至 10.5.0.21 以及 1
Hackers exploit critical Adobe Commerce flaw to hijack customer accounts
CVE-2026-71362
Attempts to exploit a critical vulnerability (CVE-2026-71362) in Adobe's Commerce and Magento e-commerce platforms have
A flaw was found in open-iscsi's iscsiuio component. This vulnerability involves an integer underflow and out-of-bounds read during Dynamic Host Configuration Protocol for IPv6 ...
CVE-2026-18727
A flaw was found in open-iscsi's iscsiuio component. This vulnerability involves an integer underflow and out-of-bounds
A flaw was found in open-iscsi. This vulnerability allows a remote attacker on the same local network segment to cause a Denial of Service (DoS) in the iscsiuio daemon. By sendi...
CVE-2026-18726
A flaw was found in open-iscsi. This vulnerability allows a remote attacker on the same local network segment to cause a
Russh is a Rust SSH client & server library. Prior to 0.62.4, an unauthenticated SSH client can cause a denial of service by sending SSH_MSG_KEX_ECDH_INIT with a 32-byte all-zer...
CVE-2026-73430
Russh is a Rust SSH client & server library. Prior to 0.62.4, an unauthenticated SSH client can cause a denial of servic
Russh is a Rust SSH client & server library. Prior to 0.62.4, a malicious SSH server can crash a russh client session with a malformed KEX_ECDH_REPLY containing a server ephemer...
CVE-2026-73429
Russh is a Rust SSH client & server library. Prior to 0.62.4, a malicious SSH server can crash a russh client session wi
A flaw was found in the provider-credential-controller component of multicluster-engine (MCE). An attacker with specific permissions on the hub cluster, and knowledge of a prior...
CVE-2026-19130
A flaw was found in the provider-credential-controller component of multicluster-engine (MCE). An attacker with specific
Apache Airflow's environment-variable secrets backend resolved a team-scoped Connection or Variable from the wrong team's scope. The guard meant to prevent this only ran when no...
CVE-2026-68076
Apache Airflow's environment-variable secrets backend resolved a team-scoped Connection or Variable from the wrong team'
Vulnerability-Lookup contains an authentication weakness in its account activation and password-recovery mechanism. Activation and recovery links were generated using stateles...
CVE-2026-73431
Vulnerability-Lookup contains an authentication weakness in its account activation and password-recovery mechanism. Ac
Docker Sandboxes (sbx) applies the read-only intent of a runtime host mount to the in-guest container bind only: the underlying virtio-fs host-edge grant is added to the sandbox...
CVE-2026-18171
Docker Sandboxes (sbx) applies the read-only intent of a runtime host mount to the in-guest container bind only: the und