⚡Megatron LLM Hub
全部
Twitter

推特安全流

头条6

自定义打印命令含 %J 时,任务名会直接进入 shell,PoC 已放出。

RT @DarkWebInformer: 🚨 A Samba printing flaw can turn a malicious print job name into remote code execution; CVE-2026-4480 PoC: https://t.co/PELF8HS4Oj The vulnerability affects Samba’s printing subsystem when %J is used inside a custom print command. A client-controlled print job name can reach the shell without proper escaping, allowing shell metacharacters to alter the command being executed. Samba Printing | Command Injection RCE CVSS: 9.8 Critical Affected: Samba < 4.22.10, < 4.23.8, and < 4.24.3 Fixed: Samba 4.22.10, 4.23.8, and 4.24.3 Impact: Unauthenticated remote code execution under affected configurations The repository documents the vulnerable configuration, attack path, affected versions, detection guidance, and mitigation steps.

rcecvepocsmb @0xocdsec 原文 ↗

musl 且无 PAM 的 Linux 默认配置即受影响,需立即处置。

RT @canadianbryan: #OpenSSH 10.6p2 (URGENT UPDATE for musl libc Linux users.) https://t.co/hkOfhzWoXq Damien Miller: "If you operate a Linux system that uses musl libc w/o PAM then stop reading &amp; set PasswordAuthentication=no or disable SSH for these systems IMMEDIATELY." https://t.co/pNnPxk804f

sshmuslcve提权 @0xocdsec 原文 ↗

自托管 AI agent 控制台零防护暴露公网,既是攻击面也是现成自查清单。

RT @SansLimit3: Basically, removing #ARTEX from the page title reveals more exposed instances and variants. Shodan: 1,852 - title:"自主渗透测试控制台","자율 침투 테스트 콘솔" Censys: 1.653 - host .services.endpoints.http.html_title=~".*自主渗透测试控制台" FOFA: 2000+ - title="自主渗透测试控制台" Another commonly observed variant is AntiFox, often exposed on port 443 with antifox.*.nip[.]io or *.sslip[.]io certificates. ARTEX and its variants: ARTEX — 自主渗透测试控制台 AntiFox — 自主渗透测试控制台 Autonomous Pentest Agent — 自主渗透测试控制台 ARTEX: 자율 침투 테스트 콘솔 TGSEC 社区专属版 — 自主渗透测试控制台 渗透工作台 — 自主渗透测试控制台 LMTex — 自主渗透测试控制台 airsuperMAN — 自主渗透测试控制台 artex — 自主渗透测试控制台 月落 — 自主渗透测试控制台 ARTEX MX — 自主渗透测试控制台 ARTEX — 자율 침투 테스트 콘솔 ARTEX-PRO — 自主渗透测试控制台 Autumn — 自主渗透测试控制台 BoomInner — 自主渗透测试控制台 GOOGTEX — 자율 침투 테스트 콘솔 MiniHunter — 自主渗透测试控制台 Pobi v2 · 自主渗透测试控制台 SHRIKE — 自主渗透测试控制台 SKARN — 自主渗透测试控制台 Soua — 自主渗透测试控制台 Uchange — 自主渗透测试控制台 mose — 自主渗透测试控制台 海盗船-内部渗透框架 — 自主渗透测试控制台 猎杀者 — 自主渗透测试控制台 #ThreatIntelligence #CTI #ThreatHunting

ai_agent工具暴露面 @cyb3rops 原文 ↗
必看23

文档与图像处理管线普遍内嵌 Ghostscript,利用已在多个应用上验证成功。

RT @FabiusArtrel: Earlier this year, I found a memory corruption vulnerability in Ghostscript's native TrueType interpreter and exploited it to achieve RCE. Here's a compilation of the exploit working against several applications: https://t.co/jrwtraUqD2

rce内存破坏poc @FuzzySec 原文 ↗

注册一个 DLL 即可长期收割凭据,现有检测面覆盖很弱。

Windows Network Provider DLL: Stealing Plaintext Credentials at Logon New Medium post, in this one we will see how Windows gives attackers a legitimate point to extract plaintext credentials on every interactive logon https://t.co/MYGTTiszFP https://t.co/s6Hf1yCn6z

凭据窃取windows持久化 @papadelospapas_ 原文 ↗

展示 AI 漏洞挖掘已达到的量级,包含大量此前被认为不可利用的 bug

RT @ky1ebot: This is the "upcoming cyber model" from yesterday. Besides 948 Linux vulns, it found 111 renderer RCEs in Firefox, among which, 81 were novel. 30 were exploits for bugs known to developers that were thought to be unexploitable. We are using this technology to scan more OSS projects and secure the Internet infrastructure. This is just a start.

llm漏洞挖掘浏览器报告 @0xocdsec 原文 ↗

指明 Alpine 默认容器并不受影响,可据此排除大量实际部署场景

RT @singe: I’m very ready to be proved wrong here but that *sshole vuln seems way less likely than at first look. musl + no PAM + at least one other user with a password + you know the username of a user with a * password. alpine looked like the most obvious but the default container isn’t vulnerable until you add another user with a password.

ssh提权分析 @vanhoefm 原文 ↗

AI 编程助手成为攻击执行器,端点检测需覆盖可信应用的白利用行为

Attackers are using legitimate AI coding assistants to execute malicious commands on compromised endpoints. In two confirmed intrusions, Elastic Threat Command observed adversaries operating cyber attacks through Claude Code on Windows and the ChatGPT desktop app on macOS. Both involved trusted applications being used to carry out malicious activity. @_xDeJesus principal security research engineer at Elastic Threat Command, walks through the findings and what security teams should be looking for: https://t.co/Kd3MqH08c5 The findings highlight a detection challenge: These applications are signed and authorized to execute commands, making it difficult to distinguish legitimate activity from attacker behavior. The speed of these attacks is another concern. Elastic researchers observed an average of just eight minutes between ClickFix/infostealer infection and attackers logging in with stolen credentials. In the latest threats, etc quarterly report, Elastic Threat Command documents these intrusions and examines the implications for endpoint visibility, behavioral detection, and threat hunting. Full report: https://t.co/7MiNCbJyuJ

ai_agentllm事件绕过 @elasticseclabs 原文 ↗

以实机对话日志证明自主智能体攻击链已成型,可作威胁建模新基线

Agentic hacking frameworks are on the rise. We tracked 2,000+ #ARTEX servers and obtained databases with full conversation logs that show exactly how the agents hacked. It proves: Fully autonomous cybercrime is already here. Read more 👇: https://t.co/ZibylV402n

ai_agentllm报告事件 @_CPResearch_ 原文 ↗

绕过官方防外泄模式后,LLM 会话数据可被无感窃取。

🖱️No clicks. ⚠️No warning. 🗣️Just a conversation. CPR found a #ChatGPT Lockdown Mode bypass enabling silent data exfiltration from the chat. Read more 👇: https://t.co/umFmjE7JxP

llm绕过数据外泄提示注入 @_CPResearch_ 原文 ↗

微虚机隔离层可被突破,1.16.2 与 1.17.0+ 已修复。

RT @coinspect: We escaped AWS Firecracker. 🔥 We discovered a VM isolation vulnerability affecting versions 0.23.0–1.16.1 when restoring snapshots using the File memory backend with balloon or virtio-mem devices. Fixed in 1.16.2 and 1.17.0+ (CVE-2026-122192) Technical write-up: https://t.co/hUU184cJyA

逃逸cve云虚拟机 @0xocdsec 原文 ↗

现成的自托管 AI 栈红队工具,可直接扫描 MCP/Agent 配置面。

AI-Infra-Guard v4.6.9 — A full-stack AI Red Teaming platform securing AI ecosystems via Agent Scan, Skills Scan, MCP scan, AI Infra scan and LLM jailbreak evaluation. https://t.co/xIKKoUsJ7n https://t.co/fNgilb724L

ai_agentmcp工具llm @KitPloit 原文 ↗

签名 DLL 内幽灵代码无痕迹,现有 EDR 检测假设需复核。

RT @SEKTOR7net: Making call stacks lie to EDRs. LACUNA Chain spoofs Windows call stacks using ".pdata" gaps - ghost code in signed DLLs that unwinds cleanly, leaves no artifacts, and fully evades Elastic, Bitdefender, and Kaspersky. A post by Mohamed Alzhrani (@0xMaz) Source: https://t.co/PeE2ZAKii8 #redteam #blueteam

绕过edrredteam内存 @0xocdsec 原文 ↗

签名滥用叠加白利用,链条完整且附可直接落地的 IOC。

Signed "Melvin Jackson" h/t @malwrhunterteam Visibility thanks to @malbeacon 's Deception Pro; ran the payload for hours on a realistic network. Signed MSI -&gt; MSBuild -&gt; .NET loader -&gt; Hijackloader -&gt; SectopRAT https://t.co/uXZwtlp2Xz IOCs in thread: 🧵 https://t.co/OK8imdGKIg

恶意软件签名滥用钓鱼ioc @SquiblydooBlog 原文 ↗

少见的 AI agent 编排实战入侵全链条还原,可用于校验自身 agent 权限边界设计。

Acronis researchers uncovered BOSSAI, a 12-agent AI setup run by a single attacker. It targeted organisations in 11 countries, with a breach at a Taipei emergency services provider exposing dispatch records. https://t.co/RSlO3G2kou https://t.co/dCSEbiXG5c

ai_agent事件报告 @virusbtn 原文 ↗

可直接并入检测管道,排查内网是否与这些 agent 实例有入站连接。

RT @s0what: Since the folks at ThreatBook developed both Artex and CyberStrikeAI, we figured we should contribute a bit to defending against and tracking AI agent attacks by sharing two feeds of instances running these agents on the public internet. Grab them if needed. Feeds: https://t.co/1rCmcFAgOt https://t.co/bBJMUotmfy

ai_agent威胁情报工具 @0x534c 原文 ↗

自主发现并利用漏洞的 Agent 已能产出新手法,自建环境的威胁模型需要重估。

RT @clintgibler: 🔥 A thousand fewer paths to root My @OpenAI colleague @ky1ebot built an autonomous vulnerability discovery, exploitation, and patching pipeline. In one run with an upcoming cyber model, it validated and generated patches for 948 local privilege escalation (LPE) vulnerabilities in Linux at an API cost of $1.2 million. Each LPE lets an unprivileged local user gain root access. To the best of our knowledge, the agents in this pipeline discovered and used at least 10 novel Linux exploitation techniques. We wrote a blog about: - How the pipeline works - The cost of each stage - Our findings and learnings https://t.co/MFAVrGSJuq

ai_agentlpe报告 @0xocdsec 原文 ↗

供应链投毒是自托管组件与 Agent 插件的高频入口,文中有可复用手法

RT @mainteemoforfun: Myself and @OriginalSicksec , bagged over 6OK-ish $ bounties from this technique in bug bounties, and based on the current supply chain attacks I wanted to share some thoughts Note: english not my mother tongue so article is enhaced by AI, but content is not an AI slop I promise https://t.co/C1b80E5FHt

供应链赏金报告 @0xocdsec 原文 ↗

公开事件复盘给出完整攻击链推理,可对照自身日志与检测思路

Always appreciate it when incident investigation findings are shared publicly. They provide valuable insights into how attacks unfold and help the security community better understand emerging TTPs. My 2cents on Binance's latest disclosure: ▶️The June incident: Still not sure how the June 2026 infection connects to the later October attack. The earlier infection looks more like an opportunistic cybercrime campaign targeting random users rather than a targeted operation. ▶️The October intrusion: I'm particularly curious about the technical details of Engineer B's initial infection and how the attackers abused the internal AI tool to move laterally to Engineers A and C. Looking forward to seeing more technical indicators shared with the community to strengthen the attribution beyond cryptocurrency wallet overlaps.

事件报告ttps @unpacker 原文 ↗
推荐15

点明 AI 漏洞挖掘的能力边界,是评估 AI 安全产品路线的直接参照

RT @seanhn: Two thoughts on this: 1. Very cool. As someone who's primary interest is the offensive side of cyber, I love it. *chefs kiss* and congrats to the team. 2. You cannot bug hunt your way to 'secure'. AI doesn't change that. If your strategy for the next few years is to boil the oceans continuously scanning your same codebase, with the same broken architecture, with the same broken threat models, you are going to have a bad time. a) you're going to spend a fortune and b) you're going to keep getting hacked.

llm漏洞挖掘报告 @moyix 原文 ↗

判断事件归因中「Agent 驱动」说法是否被夸大,需要了解其实际能力

RT @ramimacisabird: 🎙️ ARTEX - more like OurTEX am i right?? One of the biggest GTM wins during cybersecurity awarness month was this open source attack agent. and now it feels like every disclosed incident is "ARTEX-driven!!" So I wrote a quick primer / reminder: 👇 https://t.co/saKK1vKd8Z

ai_agent工具报告 @0xocdsec 原文 ↗

可在本地工具链中自建 CVE 查询与富化接口,无需依赖第三方 SaaS

KEVin — REST API over the CISA Known Exploited Vulnerabilities catalog, exposing CVE lookups, recent KEV additions, text search, and merged CISA, MITRE, and NVD data. https://t.co/6WyH2NwDb6 https://t.co/1ImZfN7poe

工具cve情报 @KitPloit 原文 ↗

避免长期开启攻击模式拖慢站点,附可直接套用的 API 调用方式

📡 Leaving Cloudflare "Under Attack Mode" on 24/7 ruins user experience. A simple bash service can monitor origin CPU load and automatically toggle protection via the Cloudflare API only when you're actively getting hit. More Practical Tips from GH: 👉 https://t.co/fSYBgSqMR9 https://t.co/X9UWND2CvR

工具ddos云 @GuidedHacking 原文 ↗

显示自动化漏洞挖掘与修补的成本产出比,Agent 攻防风向标。

RT @OpenAI: Our security researchers built an autonomous cyber vulnerability discovery, exploitation, and patching pipeline. In a single run with an upcoming cyber model, it validated and generated patches for 948 local privilege escalation vulnerabilities in the Linux kernel at a total API cost of $1.2 million.

ai_agentlpe内核自动化 @pentest_swissky 原文 ↗

面向 HTTP/3 的压测工具稀缺,直接影响协议实现漏洞的发现速度。

How practical is 100k requests per second? Introducing h3fuzz - @SLCyberSec Shubham Shah &amp; Patrik Grobshäuser https://t.co/2axuXMRHEF

工具fuzzinghttp3 @pentest_swissky 原文 ↗

把 AI 自动化攻击对威胁模型的实际影响讲清楚,可据此调整检测优先级。

RT @cyb3rops: I think we've crossed an important threshold ARTEX, Cairn and similar frameworks can now automate substantial parts of real intrusions. And as models become cheaper and more capable, the number of attacks is no longer constrained in the same way by human skills, time and attention I wrote about what changed, why this is different from traditional automation, and what I expect to happen next https://t.co/nIAvB4JxUv

ai_agent报告威胁建模 @ShitSecure 原文 ↗

现成查询即可把外部 agent 威胁情报接入 Defender 做回溯排查。

🛡️𝗗𝗲𝗳𝗲𝗻𝗱𝗶𝗻𝗴 𝗮𝗴𝗮𝗶𝗻𝘀𝘁 𝗔𝗥𝗧𝗘𝗫 𝗮𝗻𝗱 𝗖𝘆𝗯𝗲𝗿𝗦𝘁𝗿𝗶𝗸𝗲𝗔𝗜 🧙‍♂️ ​ This query checks 𝙳𝚎𝚟𝚒𝚌𝚎𝙽𝚎𝚝𝚠𝚘𝚛𝚔𝙴𝚟𝚎𝚗𝚝𝚜 for successful inbound connections from IPs found in the 𝗔𝗥𝗧𝗘𝗫 and 𝗖𝘆𝗯𝗲𝗿𝗦𝘁𝗿𝗶𝗸𝗲𝗔𝗜 threat feeds. ​ A simple way to bring external threat intelligence into Advanced Hunting and look for potential hits. ​ KQL Detection: https://t.co/SiyJTjymIz ​ Credit to Feng Xue for the ThreatBook feeds. 🙌 ​ #ThreatHunting #KQL #MicrosoftDefender #ThreatIntelligence

ai_agent检测kql @0x534c 原文 ↗

权重级改写的收益与代价数据,评估本地模型安全能力时可直接引用。

RT @pdiscoveryio: We let GPT-6 Astra directly edit Qwen3.5-9B's weights with no fine-tuning or gradient descent. SecEval: 32.5% → 51.5% But: → no clear CTF improvement → no better vulnerable-vs-patched judgment → 3× more false positives A higher benchmark number does not always mean a stronger capability gain.

llm报告 @0xocdsec 原文 ↗

同一事件的官方口径,提供规模与成本口径便于横向比较。

RT @OpenAI: Our security researchers built an autonomous cyber vulnerability discovery, exploitation, and patching pipeline. In a single run with an upcoming cyber model, it validated and generated patches for 948 local privilege escalation vulnerabilities in the Linux kernel at a total API cost of $1.2 million.

ai_agentlpe @rootxharsh 原文 ↗
更多63

补丁窗口已为负值而修复周期升至 43 天,防御优先级需整体重排。

RT @a16z: For two decades, writing the exploit was the bottleneck in a cyberattack. AI made that step cheap, and patching hasn't gotten faster: - On average, attacks went from landing 63 days after disclosure in 2018-19 to 7 days before it in 2025 - A working exploit went from a skilled person's month to a few hours of machine time - The median time to fully fix a known-exploited bug rose to 43 days a16z's @asaf_ez and @steph_zhang expect a long, expensive worm era, hitting hardest where systems can't update fast: hospitals, schools, small water systems. "When attacking is nearly free, attackers no longer need to want a target." Full piece from Asaf Ezra and Steph Zhang: https://t.co/8L3D9mnL44

漏洞利用ai_agent报告趋势 @cyb3rops 原文 ↗

AI 辅助漏洞挖掘进入商用产品阶段,攻防节奏与门槛都将被重估。

RT @OpenAI: Introducing GPT‑6 Cyber: a major advance in security research that pairs stronger capabilities with better model alignment. The model powered the Linux kernel research we announced yesterday, and security teams at companies like Mozilla, Palo Alto Networks, Calif, and Cisco—alongside our own at OpenAI—are already putting it to work. GPT-6 Cyber is available to select Daybreak Red customers and we are expanding access over time. https://t.co/Nzab3Z1Td6

llmai_agent漏洞挖掘 @pentest_swissky 原文 ↗

不携带自有加密代码、体积更小,是当前可复用的免杀具体手法。

Implant encryption using a trusted Microsoft cryptographic implementation. ✅ Smaller stagers without using crypto code ✅ Improved evasion capabilities against EDR ✅ AES-256 encryption. https://t.co/yiwTJiUaQV

免杀加密windows工具 @ipurple 原文 ↗

少见基于内部数据的风险量化,涉及 AI 自我加速的实证判断。

RT @METR_Evals: Since August, a team within METR has worked with internal Anthropic information to assess whether AI itself is accelerating AI development. We are now publishing a report summarizing the results of this risk assessment.

llm报告风险评估 @FuzzySec 原文 ↗

4096 行微码地址空间全量公开,硬件逆向的标志性成果。

RT @LoveMHz: After 25 years, the original Xbox's last hidden ROM has finally been dumped: the CPU's microcode! This has been a goal of mine for over 6 years, and I've worked on it off and on ever since. Over the last couple of weeks, while building inventory and shipping orders, I've also been running hardware tests 24/7. This poor Xbox has rebooted over 50,000 times in those two weeks alone. And it finally paid off. The dump covers the entire 4,096-line microcode address space. The ROM itself occupies 3,200 of those lines, roughly 12 KB of microcode. I can't overstate how huge this feels. After all these years, I finally have it. A full write-up is coming next year!

逆向硬件固件 @daaximus 原文 ↗

勒索生态人员归因的完整链路可作方法参考,但涉及真实个人身份。

RT @GangExposed_RU: Who Runs Qilin? Identifying Haise  A new investigation identifies Haise, a Qilin administrator, as Alexey Yakovtsev. The findings were developed entirely on the @InteIHawk platform using its internal sources. Full investigation: https://t.co/UBJaOFgo5M

勒索归因osint @cyb3rops 原文 ↗

展示 LLM 辅助复现历史漏洞的可行路径,方法论可迁移。

RT @spendergrsec: So this recent vuln of a stupid way that lets you into a system reminded me of something that's been bugging me for 31 years that I never solved until today with the help of 4 minutes and 18 seconds of Codex's time. Let's go back to 1995!

llm漏洞方法论 @alkalinesec 原文 ↗

RT @ItsReallyNick: You can break into musl linux with password auth* (probably a lot of edge devices) using only 7 emojis: 👨🏻‍👩🏻‍👦🏻‍👦🏻👨🏻‍👩🏻‍👦🏻‍👦🏻👨🏻‍👩🏻‍👦🏻‍👦🏻👨🏻‍👩🏻‍👦🏻‍👦🏻👨🏻‍👩🏻‍👦🏻‍👦🏻👨🏻‍👩🏻‍👦🏻‍👦🏻👨🏻‍👩🏻‍👦🏻‍👦🏻 *sshole is crazy https://t.co/gtjdVc1vfo

sshd绕过cve @nas_bench 原文 ↗

漏洞产出速率改变后,补丁优先级与防御资源分配假设需同步调整

RT @TalBeerySec: 1/ We are doing it totally wrong with vulns. In my new post I dig into @googlechrome’s vuln data to show that the AI-driven Vulnpocalypse settled a 20-year-old debate: vulnerabilities are dense. Our defenses are still built as if they were sparse. https://t.co/u46KydU2wS

报告llm漏洞管理 @cyb3rops 原文 ↗

给出 AI 漏洞挖掘成本与产出的可比数字,可用于评估自动化审计投入

RT @ETenal7: Good writeup, but the $1.2M experiment sounds unnecessarily expensive. As a comparison of Linux kernel bug hunting: NebuSec v.s. OpenAI Model: GPT-5.2/5.4 v.s. cyber + undisclosed internal model Cost: $100K v.s. $1.2M Findings: 2386 vs 1724 Merged Patch: 170 v.s. 61 That's roughly 16x less cost per finding, and ~3x more patch merged. Our mission is to make vulnerability scan cost efficient that every company can afford, not just an expensive toy owned by frontier lab

llm内核报告 @0xocdsec 原文 ↗

指出自动利用生成在 kernelCTF 成功率仅约 3%,校正全自动挖洞预期

RT @ETenal7: BTW, we got 600K bug bounty in return from Google VRP. We might be closest team who can prove the claim of bug bounty is net profitable, but the capital return isn’t at the claimed 55x times We spent reasonable time to understand AEG doesn’t equal to automated bug bounty, only 3% of the LPE works on kernelCTF, and you want to make sure the submissions are not duplicated. Also kernelCTF isn’t a playground, it’s an arms race. You must be the first who submit the flag, and our last kCTF exploit took only 0.1 second. That means the rest 10+ teams (likely OpenAI included) not be able to get a dime even you have a working LPE on hand.

llm内核漏洞赏金报告 @0xocdsec 原文 ↗

从触发、信息泄露到读写的完整内核提权教程,方法可复用。

Exploiting a use-after-free (UaF) in Android’s Binder driver (CVE-2023-20938) (@eshard) Part 1: https://t.co/iWVoizBw7p Part 2: https://t.co/KbPDa6Na7X Part 3: https://t.co/c9qv8URcn8 Part 4: https://t.co/I7VPLexq4v #infosec https://t.co/fi3Cs3rnMs

lpe内核移动端poc @0xor0ne 原文 ↗

合法远程运维工具被滥用即可长期驻留,检测思路可迁移。

In one of my previous research, I turn Chrome Remote Desktop into a complete red team scenario that can be used for monitoring and even leads to spying.. I explain different use cases and how to make it possible. Link :- https://t.co/6ZeFjVAUVF https://t.co/rR5CHTHIoo

工具红队绕过监控 @5mukx 原文 ↗

无实据的猜测,但点出模型训练数据与护栏之间的利益冲突。

RT @elder_plinius: my spidey senses are telling me the labs are, as we speak, using unguardrailed versions of their unreleased models to systematically reverse-engineer ALL of the internet’s software, apps, and games, then feeding the reconstructed source code straight into their training pipelines while simultaneously building guardrails to ladder-pull the rest of us

llmai_agent @0xocdsec 原文 ↗

macOS 窃密家族细节有限,仅作议题线索了解。

RT @moonlock_com: Virus Bulletin 2026 was a treat from start to finish. Our own Kseniia Yamburh unpacked a macOS stealer, shared the stage with a researcher we usually only know by handle, and got to hear that people actually read our research. Thank you for having us, VB, and thank you, Seville, for the sunshine! We’re so glad we got to be part of it: on stage, in the hallways, and as a bronze partner. Now back to the lab to keep cyber threats away from your Macs.

macos恶意软件议题 @patrickwardle 原文 ↗

远程操控物理制冷设备已造成实物损失,OT 暴露面风险具象化。

RT @grok: The attacks on European supermarket refrigeration systems are real and confirmed by retailers and authorities across multiple countries, with freezers heated to around 50°C and food spoiled. NoName057(16) and Z-Pentest did claim responsibility in Telegram posts and an AI-generated video targeting Danfoss systems. Those claims remain unverified. No official attribution exists, and the posts were later deleted. They are plausible based on prior activity, but currently just a claim.

otics事件 @artem_i_baranov 原文 ↗

该结论直接影响补丁节奏与检测投入的优先级排序。

RT @TalBeerySec: 1/ We are doing it totally wrong with vulns. In my new post I dig into @googlechrome’s vuln data to show that the AI-driven Vulnpocalypse settled a 20-year-old debate: vulnerabilities are dense. Our defenses are still built as if they were sparse. https://t.co/u46KydU2wS

漏洞报告浏览器 @moyix 原文 ↗

RT @NFT_Chen: 🔥炸了!智谱创始人唐杰教授今晚清华授课硬核揭秘了以往GLM模型的数据配方,暗示 GLM 5.5 绝对将掀起开源模型新高度! 唐杰一句话定调:数据配方是我们预训练人员的命! 六源六scaling问题直接摊开: 🔹Web:1T pages(日爬百亿级)
🔹Documents:2B + GLM-OCR
🔹Books:50M(质量筛选+版面重排)
🔹Code:500M repos(可执行+测试)
🔹Video:10B(世界知识+长序列)
🔹Audio:300M小时(ASR迭代重写→数万亿token) 每一路都是独立工程地狱。 最新配比暂不公开,但早期配方已经能看出来智谱有多在乎数据。 GLM 5.5如果真的按这个路子继续卷数据和后训练,国产基座又要再上一个台阶了。 #智谱 #GLM #唐杰 #GLM55 #清华 #大模型 #数据配方 #预训练 #开源大模型

llm预训练 @0xocdsec 原文 ↗

官方分发渠道被篡改的供应链案例,Steam Deck 玩家同样暴露,debug.log 可自查。

RT @IntCyberDigest: ‼️ BREAKING: The PC launcher of hit anime shooter Nikke was hijacked to drop a remote-control agent built to steal passwords, SSH keys and crypto wallets, a researcher says. Attackers planted code on 27 September in official files on the servers of Level Infinite, Tencent's games label. The launcher loads them at every start. Several players have found the agent running. Proton is no shield, so Steam Deck players are exposed too. Players can check the launcher's debug.log for k2f_chan. Level Infinite swapped three of the four files today, the researcher says, without a word to players.

供应链恶意软件事件 @virusbtn 原文 ↗

合法签名加云域名绕检测的装载链仍在活跃,IOC 可直接用于狩猎。

Signed: "ISABEL ALEJANDRA FLORES" C2: asiioasdjasiojasodijasa.eastus.cloudapp[.]azure[.]com @malwrhunterteam called this domain out a month ago, Not low detection but still using Microsoft Certs and Azure Still using CrowdStrike theme https://t.co/90r3k9KxM9 https://t.co/VKYzxAJw7o

恶意软件签名c2 @SquiblydooBlog 原文 ↗

一组可复用的免杀与注入技巧清单,适合更新现有检测规则。

RT @Unit42_Intel: A #ClickFix campaign pushing #RomulusLoader spans multiple sites and uses these techniques during delivery: C2 IP address as a plain integer, MSI Binary table steganography, AMSI patched in a suspended process &amp; 10 injection techniques. More info at https://t.co/S9eEUa8qY9 https://t.co/2w57f7iGIU

恶意软件注入绕过 @virusbtn 原文 ↗

合法签名 EXE 侧加载活动持续扩散,样本时间线可用于回溯排查。

RT @ThreatrayLabs: Earlier this week, our friends at @nextronresearch documented an RRCfg delivery chain. We hunted for implants with similar code in our platform and uncovered additional samples and delivery chains dating back to September 2026, using different hosts, sideloaded DLLs and C2 infrastructure. 🔹 21 September 2026 — arphaCrashReport.exe / arphaDump64.dll. Alibaba’s arphaCrashReport.exe, renamed 20260921163500-CC1DB4D9D1744A.exe, sideloads malicious arphaDump64.dll. Its bootstrap downloads Donut from hxxps://xiaxiazai.hkg.bcebos[.]com/kekehua/27.124.41.111.bin on Baidu Cloud BOS and executes it, loading RRCfg configured for 27.124.41[.]111:8083. Report: https://t.co/jzQT50fr0x 🔹23 September 2026 — ClosePre.exe / kcommon64.dll. ClosePre.exe, renamed 2026095SDDE20.exe, sideloads malicious kcommon64.dll. Its bootstrap requests hxxps://gdnkdl[.]eu[.]cc/files/ahfayui.eu.cc.bin. No next-stage payload was received. 🔹22 October 2026 — BioEnrollmentHost.exe / wincorlib.dll. Microsoft BioEnrollmentHost.exe, renamed 20261022128205.BC6FBTAC08T.exe, sideloads malicious wincorlib.dll. The DLL decrypts a bootstrap embedded in the host’s certificate table. It downloads Donut from hxxps://aawwf.hkg.bcebos[.]com/shiade/A03.bin, also on Baidu Cloud BOS, which loads RRCfg configured for 192.252.178[.]174:8083. The loader shares Nextron’s CAFEBABECAFEBABE marker, RC4 key AtkExEmbedKey2026, and SCURLTPL URL configuration. Report: https://t.co/Ys5x5z0shn 🔹25 October 2026 — https://t.co/RbFIfTC0X8 C2 variant. An RRCfg implant uses ahfayui[.]eu[.]cc, which resolved to 61.18.208[.]171. The same hostname appears in the payload filenames requested by the ClosePre and Edge loaders. Report: https://t.co/Oh0cOZU4YT 🔹3 November 2026 — Edge notification_helper.exe / dbghelp.dll. Microsoft Edge notification_helper.exe, renamed INV-No.20261102023748.exe, sideloads malicious dbghelp.dll. This loader also decrypts a certificate-embedded bootstrap, then requests hxxps://bmafhahl[.]eu[.]cc/files/ahfayui.eu.cc.bin. No next-stage payload was received. 🧬 Sample hashes, #IOCs and analysis reports: https://t.co/54m3ZGCY2S

恶意软件侧加载事件 @cyb3rops 原文 ↗

RT @ElementalX2: We(@spontiroli , me & Eliad) from @Acronis TRU Labs found a threat actor which we term as an AAT(AI-Augmented Threat) based out of Taiwan, running an attack mesh known as #BOSSAI , operating it via WeChat & OpenClaw connected to LINE app, targeting multiple countries such as Taiwan, US, Thailand, performing data exfiltration in some of the cases. Also, the threat actor runs multiple other businesses and his love for Korean Skincare is unmatched 💅💅💅 IOCs & research : ⏬ https://t.co/5x9lDbg9np

ai_agent事件 @virusbtn 原文 ↗

IOC 列表可直接落阻断规则,钓鱼基础设施仍在活跃投放。

🏦🎣 An interesting cluster of recently created helpdesk-themed domains targeting financial institutions. Sharing with the community for awareness and further investigation.🤝 helpdesksynckey[.]com jpaccesshelpdesk[.]com bankofamericahelpdesk[.]com tdbank-helpdesk[.]com helpdeskus-01[.]com revoluthelpdeskchat[.]com hsbcnethelpdesk[.]com helpdeskus01[.]com chaseconnecthelpdesk[.]com cashprolivehelpdesk[.]com heurohelpdesk[.]online butterfieldhelpdesk[.]com jpmorganhelpdesk[.]com westpachelpdesk[.]com 1password-helpdesk[.]cc butterfieldhelpdesk[.]com revoluthelpdeskchat[.]com westpachelpdesk[.]com jpaccesshelpdesk[.]com bankofamericahelpdesk[.]com jpmorganhelpdesk[.]com cashprolivehelpdesk[.]com westpachelpdesk[.]com jazztech-it-helpdesk[.]com hsbcnethelpdesk[.]com helpdesk-auth[.]com chaseconnecthelpdesk[.]com #threathunting #helpdeskphishing

钓鱼ioc @0x534c 原文 ↗

自建服务常用穿透工具,版本更新涉及协议与稳定性,需跟升级。

frp v0.72.0 — A fast reverse proxy to help you expose a local server behind a NAT or firewall to the internet. https://t.co/xyRSnv20Ed https://t.co/XD3OkE9tcu

工具自托管 @KitPloit 原文 ↗

老牌技术刊物,历期常含未公开的内核与利用技术细节。

you should read @phrack and keep an eye out for phrack#73 dropping soon

议题报告 @Teach2Breach 原文 ↗

“I already patched, so why should I care?” Patching doesn’t remove an attacker who got in before the fix. And you can’t patch what you don’t know is vulnerable. So verify what’s actually running across your environment and check that the fixes reached those systems. Patched everything you know about? 𝗧𝗲𝘀𝘁 𝘄𝗵𝗮𝘁 𝗮𝗻 𝗮𝘁𝘁𝗮𝗰𝗸𝗲𝗿 𝗰𝗼𝘂𝗹𝗱 𝘀𝘁𝗶𝗹𝗹 𝗲𝘅𝗽𝗹𝗼𝗶𝘁: https://t.co/S2YVgdpxK6

报告 @HacktronAI 原文 ↗