make smarter security investments and mitigate future risks
make smarter security investments and mitigate future risks
2026-09-06 归档
今日共收录 918 条安全情报，包含 147 个 CVE，434 篇安全通告，以及 81 篇研究论文。
← 返回归档列表Inside Astaroth's New Spambot Component
Inside Astaroth's New Spambot Component
Falcon Platform IOAs Arrive in Falcon Next-Gen SIEM to Identify New Threats
Falcon Platform IOAs Arrive in Falcon Next-Gen SIEM to Identify New Threats
CrowdStrike 2026 Threat Hunting Report: Exploitation Window Closes as AI Use Accelerates
CrowdStrike 2026 Threat Hunting Report: Exploitation Window Closes as AI Use Accelerates
CrowdStrike Threat Hunts for Shell Command Obfuscation on VMware ESX
CrowdStrike Threat Hunts for Shell Command Obfuscation on VMware ESX
Google is a Leader in the IDC MarketScape: Worldwide Cybersecurity Consulting Services 2024 Vendor Assessment
Google is a Leader in the IDC MarketScape: Worldwide Cybersecurity Consulting Services 2024 Vendor Assessment
Teaching AI to Reason Through Detection Triage
Teaching AI to Reason Through Detection Triage
CrowdStrike Falcon Guardian Defines the Next Generation of AI SecuritySep 01, 2026
CrowdStrike Falcon Guardian Defines the Next Generation of AI SecuritySep 01, 2026
Peer Pressure: Inside the Sality Botnet Disruption Operation
Peer Pressure: Inside the Sality Botnet Disruption Operation
CrowdStrike Delivers the Next Evolution of the Agentic SOC
CrowdStrike Delivers the Next Evolution of the Agentic SOC
Over 5,400 hacked sites serve ClickFix payloads stored on the blockchain
A massive cybercriminal operation is leveraging thousands of compromised small-business websites to deliver ClickFix pay
OpenAI admits it didn't disclose rogue AI wiki hijacking incident
OpenAI admits it did not disclose an incident where autonomous AI agents hijacked a German wiki, created 18,000 posts, s
Can Risk-Based Alerting Mitigate Cybersecurity Alert Fatigue?
本文针对安全运营中心（SOC）面临的网络安全告警疲劳问题，首次系统性地评估了基于风险告警（Risk-Based Alerting, RBA）方法的有效性。SOC 日常需要处理海量告警，其中大量为误报，导致真实攻击在有限的人力资源下难以被及时
The Implications of Linguistic Illegibility for LLM Security
该论文由安全研究员 James Mickens 撰写，讨论了大型语言模型（LLM）安全中一个被忽视的根本性问题——“语言不可读性”（linguistic illegibility）。论文指出，LLM 被训练输出自然语言，但其内部计算本质上是
Counter-GEO-Bench: Evaluating Defenses Against Information-Distorting Generative Engine Optimization
生成式引擎优化（GEO）是一种通过操纵网页内容来提升其在生成式搜索引擎中可见度的技术，类似针对传统搜索引擎的SEO。然而，该技术可被攻击者滥用：他们制作看似普通但经过特殊优化的文档，当受攻击的大语言模型（LLM）在检索增强生成过程中获取并综
Inferring Hidden User Models from the Behavior of Personalized LLM Agents
该论文研究了个性化大语言模型（LLM）智能体的隐私泄露风险。随着LLM智能体将用户记忆转化为压缩或结构化表示（称为用户模型）以指导后续决策，尽管源文本在普通接口可达的状态中被移除，此类模型仍被视为具有更好的隐私保护性，因为直接的内存提取攻击
AI-Assisted Design of a Post-Quantum Cryptographic Accelerator: A Deployed-Silicon Case Study
该论文针对后量子密码（PQC）迁移中硬件实现的安全验证盲区展开研究。由于硅片一旦流片后无法远程修补，标准验收测试（如已知答案测试 KAT）的覆盖率不足可能带来致命缺陷。作者指出，在 ML-DSA 签名算法中，拒绝采样循环会根据输入消息动态调
A Blind Trust, the Bloody Thrust: When Attacker-Controlled Hook Updates Steer AI Agent Harnesses towards Malicious Behaviors
现代AI agent harness（智能体框架）通过“生命周期钩子”机制将Shell命令绑定到会话开始、工具调用和文件编辑等运行时事件。这些命令通常以宿主主机权限执行，但它们作为钩子配置分发，且可能在大型语言模型（LLM）从未感知的时刻被
Flip, Don't Shuffle: Watermarking LLMs at the Speed of Inference
该论文提出了一种名为 Stateless Bernoulli Watermarking（SBW）的新型大语言模型（LLM）文本水印方法，旨在以极低的计算开销实现可靠的模型输出标记与检测。与现有方法如 KGW（基于词汇置换和固定大小绿名单）和
AlcaTRAz - Anchored Tree-Rule Defense Against Jailbreaks
该论文针对大型语言模型（LLM）面临的越狱攻击问题，提出了一种名为 AlcaTRAz（Anchored Tree-Rule defense Against jailbreaks）的提示词级别防御方法。越狱攻击通过精心构造的输入绕过模型的安全