Microsoft Defender wrongly flags DigiCert certs as Trojan:Win32/Cerdigent.A!dha
Microsoft Defender is detecting legitimate DigiCert root certificates as Trojan:Win32/Cerdigent.A!dha, resulting in wide
2026-05-05 归档
今日共收录 918 条安全情报，包含 208 个 CVE，383 篇安全通告，以及 47 篇研究论文。
← 返回归档列表Breaking the code: Multi-stage ‘code of conduct’ phishing campaign leads to AiTM token compromise
Microsoft Defender Research observed a large-scale credential theft campaign that exemplifies this trend, using code of
CrowdStrike Technical Risk Assessments Reveal Common Exposure Patterns
CrowdStrike Technical Risk Assessments Reveal Common Exposure Patterns
Amazon SES increasingly abused in phishing to evade detection
The Amazon Simple Email Service (SES) is being increasingly abused to send convincing phishing emails that can bypass st
Trellix discloses data breach after source code repository hack
Cybersecurity firm Trellix disclosed a data breach after attackers gained access to "a portion" of its source code repos
They don’t hack, they borrow: How fraudsters target credit unions
Fraudsters aren't hacking credit unions, they are exploiting normal business processes. Flare reveals how structured loa
Progress warns of critical MOVEit Automation auth bypass flaw
Progress Software warned customers to patch a critical authentication bypass vulnerability in its MOVEit Automation ente
Webinar: Why MSPs must rethink security and backup strategies
Security breaches don't just test your defenses—they test your recovery. Join Kaseya in our upcoming webinar to learn ho
CISA says ‘Copy Fail’ flaw now exploited to root Linux systems
CISA has warned that threat actors have started exploiting the "Copy Fail" Linux security vulnerability in the wild, one
Microsoft confirms April Windows updates cause backup failures
Microsoft has confirmed that the April 2026 security updates are causing failures in third-party backup applications usi
Cisco Moves to Acquire Astrix Security to Tackle Non-Human Identity Risks
The acquisition strengthens Cisco’s push into identity-centric security for AI and machine access. The post Cisco Moves
Exploitation of ‘Copy Fail’ Linux Vulnerability Begins
CISA has added the bug to its KEV list, and Microsoft has observed limited exploitation, mainly associated with PoC test
Autonomous LLM Agent Worms: Cross-Platform Propagation, Automated Discovery and Temporal Re-Entry Defense
该论文提出并系统分析了自主LLM代理（如长期运行进程）面临的新型蠕虫传播威胁。LLM代理通常拥有持久化工作空间、内存文件、定时任务和消息集成等特性，这使得攻击者可以通过将恶意内容写入持久化状态（如文件），利用代理的定时自动加载机制使该内容重
FunFuzz: An LLM-Powered Evolutionary Fuzzing Framework
本文提出 FunFuzz，一个基于大型语言模型（LLM）的进化模糊测试框架，旨在解决传统 LLM 驱动模糊测试中 prompt 初始化和采样方差导致的探索效率低下和输入冗余问题。FunFuzz 采用多岛进化算法，并行运行多个隔离的搜索过程，
ContextualJailbreak: Evolutionary Red-Teaming via Simulated Conversational Priming
本文提出了一种名为 ContextualJailbreak 的黑盒红队攻击策略，用于自动化发现大语言模型（LLM）的越狱漏洞。与大多数现有工作集中在单轮提示优化不同，本文聚焦于多轮对话中的上下文预置（contextual priming）攻
APIOT: Autonomous Vulnerability Management Across Bare-Metal Industrial OT Networks
本文提出APIOT（Autonomous Purple-teaming for Industrial OT），首个利用大语言模型（LLM）在裸机工业OT网络中进行自主攻击与修复的框架。传统自动化渗透测试主要针对Linux和Web系统，这些系
On the Privacy of LLMs: An Ablation Study
大型语言模型（LLMs）在交互式和检索增强场景中日益部署，引发了显著的隐私担忧。以往的研究通常孤立地分析成员推断攻击（MIA）、属性推断攻击（AIA）、数据提取攻击（DEA）和后门攻击（BA），缺乏对它们在常见系统因素下行为的统一理解。本文
What's on Your Mind? Exploring Privacy of Mental Health Apps
本文对25款流行的Android心理健康与生活教练应用进行了全面的隐私分析。研究方法结合了静态分析（检查APK中的权限声明和追踪器SDK）、动态网络流量捕获（识别实际数据传输）以及基于LLM的隐私政策提取（辅以人工标注）。主要发现包括：1）
Trojan Hippo: Weaponizing Agent Memory for Data Exfiltration
本文研究了LLM代理（Agent）长期记忆系统中的安全漏洞。尽管记忆系统允许无状态的LLM代理跨会话持久化用户信息，但这也引入了新的攻击面。作者提出了一类名为“Trojan Hippo”的持久性记忆攻击，该攻击在比以往记忆投毒研究更现实的威
When Alignment Isn't Enough: Response-Path Attacks on LLM Agents
该论文首次提出并形式化了一种针对BYOK（自带密钥）智能体架构的新威胁——对齐后篡改攻击。在BYOK架构中，用户将LLM流量路由通过第三方中继，但中继在LLM生成响应后、智能体执行前可以修改结果，从而破坏对齐。作者将该威胁实例化为中继篡改攻