⚡Megatron LLM Hub
全部
Twitter

推特安全流

头条6

开源实现 hypervisor 级 shellcode 隔离,免杀与沙箱设计可直接上手改造。

RT @papadelospapas_: HyperLoader now in GitHub! Hypervisor isolated shellcode execution environment. Maps host process memory into a guest vCPU, intercepts DLL execute faults, and proxies API calls to the host, Stardust PIC shellcode compatible https://t.co/Gp625UqgGP https://t.co/IZkrFzwAFn

工具shellcode逃避 @ShitSecure 原文 ↗

打印机协议变通用认证强制入口,示例脚本随 PR 一并放出

RT @abdo_mhanni: You can now coerce HTTP authentication from Windows targets that are running MS-RPRN, using impacket with PR #2361! The PR added a new opnum from the MS-RPRN spec, OPNUM 88 and an example script to make the process straight forward. Letting you now coerce and relay HTTP authentication from remote machine targets Shoutout to sudonoodle for the BOF that inspired this and for the technique(at least I think it was them who found it) :3

强制认证ntlmimpacket工具 @ShitSecure 原文 ↗

MCP 工具链编排的完整开源实现,可在本地部署直接试用。

VulnClaw — Based on AI Agent + MCP toolchain + penetration skill orchestration, combined with large language models, natural language input → automatically completes the full process of "information gathering → vulnerability discovery → vulnerability... https://t.co/tCnFDLCEqv https://t.co/IjO1TYY7Cr

ai_agentmcp工具渗透 @KitPloit 原文 ↗
必看30

隔离边界被突破,多租户云与自建 microVM 面临跨租户逃逸,1.16.2 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

逃逸云虚拟化cve @lcfr_hax 原文 ↗

漏洞发现—利用—修补闭环已工程化,API 成本可算,攻防节奏基线被抬高。

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内核报告 @FuzzySec 原文 ↗

不只是找洞,还产出了新的利用原语,说明 agent 已能拓展攻击知识边界。

RT @ky1ebot: We built a vulnerability research pipeline, using an upcoming cyber model, to analyze the security of the Linux kernel. Besides finding and exploiting the found vulnerabilities, the agents discovered at least 10 new exploitation techniques that were new to us and some external kernel exploit experts we checked with. More on the design and what we learned: https://t.co/SGZXwSRwxU

ai_agent内核提权议题 @S1r1u5_ 原文 ↗

攻击量不再受人力与技能约束,防守方需按机器速度重排检测优先级。

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报告议题 @cyb3rops 原文 ↗

暴露面仅一个开放端口,远程可利用门槛极低。

RT @mattiasgeniar: 4/ On the target, you need one thing: any IPv4 TCP port that accepts connections. No MPTCP needed for the crash. I turned MPTCP off. It crashed anyway. The MPTCP is for the remote-code-execution proof of concept, but not limited to just MPTCP!

内核rcepoc @0xocdsec 原文 ↗

Firecracker 是多云隔离底座,guest 到宿主逃逸一旦成立影响面极大。

RT @julianor: JP (@_JcryPto_ ) escaped AWS Firecracker from guest userspace. 🧨🔥 The funniest part? While guiding/fighting the agents, he had to make one read his own 20-year-old Phrack article to teach it how to hack. Full report coming soon. Proud of the @Coinspect team!

逃逸云虚拟化事件 @0xocdsec 原文 ↗

给出下季度最可能踩到的攻击面排序,可直接用于防御优先级排期。

CyberSec Trends Q4/26 🔮 ⬆️ AI-driven attacks (ARTEX, Cairn) ⬆️ 0day / critical RCE disclosures ⬆️ Software supply-chain attacks ⬆️ Token/OAuth app abuse & cloud persistence ↗️ EDR blinding / killing (vulnerable or malicious signed drivers) ↗️ Persistence on systems outside EDR coverage ↗️ Abuse of legit RMM / tunneling tools

报告ai_agent供应链云 @cyb3rops 原文 ↗

三段完整利用链公开,是研究移动端零点击攻击面的现成样本。

RT @0xor0ne: 0-click exploit chain targeting Pixel 9, from RCE in mediacodec to kernel LPE Part 1: https://t.co/7Vx0Mmmj5i Part 2: https://t.co/du2138bdct Part 3: https://t.co/ogXwfrFDpW Credits @natashenka and @__sethJenkins #infosec https://t.co/84vxEUeBY5

rcelpe移动端poc @0xocdsec 原文 ↗

公开可得的 AI 攻击工具已被证实用于真实金融入侵,门槛显著降低。

RT @choblin29: Holy Shit. In a new critical advisory issued TODAY, India's cyber agency CERT-In says hackers are using AI-assisted tools to attack financial institutions, exploiting payment APIs and testing transfers with just ₹1 before moving larger sums. CERT-In identifies the group as CRAFTY FALCON and says it is using FOUR publicly available AI offensive-security systems: ARTEX, PentAGI, Cairn and CyberStrikeAI. The campaign is still succeeding against fintechs and NBFCs, including firms that had already been warned and attempted remediation. Some were deliberately targeted AGAIN, while attackers are timing intrusions around weekends and holidays when security monitoring is lower.

ai_agentllm事件报告 @cyb3rops 原文 ↗

一小时拿域控,AI 代理把常规手法压缩到分钟级,检测窗口随之收窄。

RT @mgreen27: Interesting post from Kaspersky on CyberStrikeAI in real intrusions. No new techniques, just speed: creds, SMB/WMI lateral movement and domain compromise in under an hour. Samples linked via Go build metadata, TLS certs and an X-Implant-Token header. https://t.co/PjLM5KS3zk

ai_agent提权报告事件 @cyb3rops 原文 ↗

主流 EDR 厂商确认 AI 客户端已进入攻击链父进程位置,本地部署同样适用。

Six observations in the new 𝘵𝘩𝘳𝘦𝘢𝘵𝘴, 𝘦𝘵𝘤 report, all of them on software you already trust. • AI clients as execution parents • A worm with npm publish rights • ClickFix on macOS • ROPC past Conditional Access • Blockchain C2 in commodity loaders • 8 minutes from stealer to login Access the full report: https://t.co/kl1ejVpmMo

ai_agent供应链报告 @elasticseclabs 原文 ↗

应用凭据长期有效且绕过条件访问,M365/Azure 环境现成攻击面。

🚨 𝗔𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗰𝗿𝗲𝗱𝗲𝗻𝘁𝗶𝗮𝗹 𝗺𝗶𝘀𝘂𝘀𝗲 𝗼𝗯𝘀𝗲𝗿𝘃𝗲𝗱 𝗮𝗰𝗿𝗼𝘀𝘀 𝟱𝟯𝟯 𝘁𝗲𝗻𝗮𝗻𝘁𝘀 ​ Earlier reporting from Todyl highlighted suspected misuse of application credentials affecting MSP Microsoft 365 and Azure environments. https://t.co/gwyK1CFBHc​ Since November 1, an operator has reportedly been authenticating 𝗮𝘀 𝗹𝗲𝗴𝗶𝘁𝗶𝗺𝗮𝘁𝗲 𝗮𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 𝘂𝘀𝗶𝗻𝗴 𝘀𝘁𝗼𝗹𝗲𝗻 𝗰𝗿𝗲𝗱𝗲𝗻𝘁𝗶𝗮𝗹𝘀 𝘁𝗵𝗮𝘁 𝗿𝗲𝗺𝗮𝗶𝗻 𝘃𝗮𝗹𝗶𝗱. ​ Initial observations: • ~𝟰𝟱 𝗮𝗽𝗽 𝗿𝗲𝗴𝗶𝘀𝘁𝗿𝗮𝘁𝗶𝗼𝗻𝘀 • ~𝟭𝟴 𝗼𝗿𝗴𝗮𝗻𝗶𝘇𝗮𝘁𝗶𝗼𝗻𝘀 𝘁𝗮𝗿𝗴𝗲𝘁𝗲𝗱 • 𝟯𝟯+ 𝘀𝘂𝗰𝗰𝗲𝘀𝘀𝗳𝘂𝗹 𝗮𝘂𝘁𝗵𝗲𝗻𝘁𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 ​ Microsoft has now reported activity dating back to 𝗢𝗰𝘁𝗼𝗯𝗲𝗿 𝟮𝟴, 𝟮𝟬𝟮𝟲. ​ Across two observed IP addresses: 🔎 𝟵𝟰𝟬+ 𝗱𝗶𝘀𝘁𝗶𝗻𝗰𝘁 𝗮𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗜𝗗𝘀 🔎 𝟱𝟯𝟯 𝘁𝗲𝗻𝗮𝗻𝘁𝘀 ​ Notably, in some cases, the 𝘀𝗮𝗺𝗲 𝗰𝗿𝗲𝗱𝗲𝗻𝘁𝗶𝗮𝗹 𝗮𝘀𝘀𝗼𝗰𝗶𝗮𝘁𝗲𝗱 𝘄𝗶𝘁𝗵 𝗮 𝗺𝘂𝗹𝘁𝗶𝘁𝗲𝗻𝗮𝗻𝘁 𝗮𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝘄𝗮𝘀 𝘂𝘀𝗲𝗱 𝗮𝗰𝗿𝗼𝘀𝘀 𝗺𝘂𝗹𝘁𝗶𝗽𝗹𝗲 𝘁𝗲𝗻𝗮𝗻𝘁𝘀 𝗳𝗿𝗼𝗺 𝘁𝗵𝗲 𝘀𝗮𝗺𝗲 𝗜𝗣 𝗮𝗱𝗱𝗿𝗲𝘀𝘀. ​ If you operate Microsoft 365 or Azure, especially in an MSP environment, this is worth hunting for. ​ 👇 Analyst report and hunting guidance below. Link: https://t.co/8GnMUYTMUq

云凭据滥用事件绕过 @0x534c 原文 ↗

给出可落地的 DLL 侧加载样本与基础设施,能直接转成检测规则。

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

供应链侧加载事件报告 @cod3nym 原文 ↗

本地自托管 Agent 拉取镜像即交出 shell,Agent 供应链投毒已真实落地

RT @NaoSouOGabe: Resolvi dar uma olhada no Artex, a "IA que hackeia sozinha" e um rumor sobre ela ter um backdor, ARTEX (campeão do desafio "Agent+" da Baidu) está sendo distribuída por um fork com 1.325 estrelas que puxa uma imagem Docker com backdoor. Quem subiu, entregou shell. 🧵

ai_agent供应链后门docker @cyb3rops 原文 ↗

三条并行的投递路径加上移动端 n-day 利用,是可复用的水坑攻击范式。

Sekoia TDR researchers identified a large-scale watering hole campaign using three infection vectors: compromised legitimate websites, Adversary-in-the-Middle-like impersonation of high-profile websites, & a compromise of a Ukrainian web analytics service. https://t.co/bBviBUMv6d https://t.co/cJmWJpxeT8

apt水坑移动端n-day @virusbtn 原文 ↗

临时 root 绕开完整性校验与云控,对移动端攻防与设备自控都是新思路。

RT @aoxzh: CVE-2026-43499 和 CVE-2026-98260 安卓手机之光,让手机控制权重回人民,什么云控什么反诈什么温控,root 后还有你说了算的事?🤪 临时 root 这招是真的天才,也不用担心在 root 后的手机打不开银行或者丢失 DRM L1 了 https://t.co/uzFtLAdWGJ

移动端cve提权绕过 @0xocdsec 原文 ↗

无需用户交互即可远程执行,影响 Android 14 至 17,11 月补丁已修复。

RT @DarkWebInformer: 🚨 Critical zero-click Android RCE patched; CVE-2026-58868 A heap buffer overflow in Android’s Framework component could allow remote code execution without additional privileges or user interaction. The flaw resides in the onOutputReceived function within FrameDecoder.cpp, where an out-of-bounds write can occur while processing media frames. • Severity: Critical • CVSS: Not yet assigned • Affected: Android 14, 15, 16, 16 QPR2 and 17 • Fixed: Security patch level 2026-11-01 or later • PoC: No public exploit identified • Exploitation: No confirmed in-the-wild attacks Administrators should aim for the 2026-11-05 patch level or later, which includes this fix and the remaining November security updates for kernel and hardware components. Source: https://t.co/oLvtMH5ntV

rcecve移动端0click @0xocdsec 原文 ↗

WordPress 部署面极广,且修复动作与触发路径重合,升级顺序需谨慎。

RT @taisic_: I found & reported a #WordPress core pre-auth RCE (triggered during the update process) with @xint_official . I submitted it as a 0-day, but WordPress released a patch almost as soon as the reporting platform confirmed the vulnerability was valid. Missed the credit, but huge shoutout to Umut Bayram for reporting first! 🙌 Plot twist: updating to the fix can expose RCE. Here's why + safe update guide: https://t.co/JWi703p6Qc

rcecveweb补丁绕过 @0xocdsec 原文 ↗

本地用户即可提权,公开利用代码已放出,未打补丁主机可直接复现。

RT @DarkWebInformer: CVE-2026-97798: Windows Error Reporting is vulnerable to command injection due to improper handling of argument delimiters, allowing an authorized attacker with local access to exploit the flaw and elevate their privileges on the affected system. Exploit: https://t.co/8eGAQfXww5 https://t.co/Ngcb1QqAUs

lpecvewindowspoc @0xocdsec 原文 ↗

手法不新、胜在速度,并给出 Go 构建信息、TLS 证书等可用的样本溯源线索。

RT @mgreen27: Interesting post from Kaspersky on CyberStrikeAI in real intrusions. No new techniques, just speed: creds, SMB/WMI lateral movement and domain compromise in under an hour. Samples linked via Go build metadata, TLS certs and an X-Implant-Token header. https://t.co/PjLM5KS3zk

ai_agent事件报告横移 @0xocdsec 原文 ↗

自托管 AI 攻击工具镜像自带后门,部署即可能失陷,需核查供应链。

RT @GrahamHelton3: Hey so if you're a red team and have been asked about ARTEX and it's capabilities based on recent news of "zomg we got hacked by advanced AI swarms using ARTEX!!!" Please take a look at this. Looks like unauth backdoor in the docker image https://t.co/WUYM0j9flO

ai_agent供应链后门docker @Teach2Breach 原文 ↗

iOS 内核 r/w 完整利用即将公开,覆盖机型广、无可回避。

RT @pwnlambda: Also saw everyone is losing their minds over the iOS kernel r/w. I’m home now and aiming to release the exploit + writeup within the next week or so. Thank you for your attention to this matter (and all the “wen eta”s) 🙏

ios内核提权poc @0xocdsec 原文 ↗
推荐15

现成脚本即可复现认证强制与中继,AD 环境横向成本降低。

RT @abdo_mhanni: You can now coerce HTTP authentication from Windows targets that are running MS-RPRN, using impacket with PR #2361! The PR added a new opnum from the MS-RPRN spec, OPNUM 88 and an example script to make the process straight forward. Letting you now coerce and relay HTTP authentication from remote machine targets Shoutout to sudonoodle for the BOF that inspired this and for the technique(at least I think it was them who found it) :3

ad工具提权 @0x64616e 原文 ↗

厂商把 rogue agent 当独立威胁面做产品,是 Agent 安全投入的风向标。

RT @CRN: CrowdStrike debuted its next set of major moves to advance the AI-powered capabilities of cyber defenders this week amid increasing urgency around protecting against fast-moving attacks that are using the same agentic-driven technologies. Key announcements from the cybersecurity giant at its Fal.Con 2026 Europe conference included the unveiling of two AI security acquisitions: @dreadnode and @fiddler_ai, along with the launch of new capabilities including controls for rogue agents and detection for AI-powered insider threats. Notably, @CrowdStrike said that the planned acquisitions and new capabilities will provide a boost to the company’s SafeMind frontier AI models and harnesses, which were unveiled at Fal.Con 2026 in September. @CRN cybersecurity reporter and Senior Editor @KyleAlspach details CrowdStrike’s five biggest moves at Fal.Con 2026 Europe here: https://t.co/zEPJPML4ga

ai_agent事件企业 @dreadnode 原文 ↗

拿到代理的工作笔记,等同于拿到攻击者学习曲线与狩猎线索。

RT @ProferoSec: Bookmark 🚨🚨 What does an AI attacker write down after it fails? We know because six operators left their agents’ notes exposed, and we analyzed them. The lessons from one victim become skills for the next. Thank you, @hackerseye13, for the collaboration. Lots of hunting ideas are waiting for you in this blog post: https://t.co/5PRjOc9Nfx

ai_agent报告事件 @cyb3rops 原文 ↗

新增机器账户强制认证原语,可直接接入中继链,PoC 已开源。

A Beacon Object File (BOF) that forces a Windows machine to authenticate over HTTP as its computer account (HOSTNAME$), by triggering a directed IPP association through the WSD port monitor. This technique allows operators to coerce and relay machine-account authentication over HTTP. https://t.co/LLoaj3m9lK

中继提权pocwindows @ipurple 原文 ↗

展示 Agent 端到端复现漏洞的门槛已降到可随手为之

RT @1leaf1bodhi: CVE-2026-43499今年7月就被披露了 让agent实现了一下没想到获得了大家的关注 贴一下相关的仓库以供参考: 漏洞发现者仓库 https://t.co/slrNFhtdnF star比较多的实用方案 https://t.co/UOpR2D74nd https://t.co/21H28NZziX 我的 https://t.co/Gm1eSwyTEY 更多的相关仓库放评论区了

ai_agentcvepoc漏洞复现 @0xocdsec 原文 ↗

把漏洞利用开发的边际成本压到小时级,防守窗口与披露节奏随之改变。

RT @_deven___: hands down the best talk I’ve ever seen. the slides are public now. - Calif got permission from @openai to discuss working with the (unreleased) “astra-cyber” model - researchers are struggling to believe some of these exploits are real - 0/1-click exploit development is possible in ~3 hours (with the right prompting, skills, and primitives) - the world is changing faster than expected I also loved the shoutout to the 2019 group facetime bugs and @natashenka’s FT work. everyone who worked on that will remember it forever ❤️ thank you so much for the presentation! @pwnlambda @brucedang @calif_io

ai_agentllm议题poc @0xocdsec 原文 ↗

现成 BYOVD 驱动可批量结束 EDR 进程,样本公开可下载,攻防双方都在用。

RT @_josehelps: LOLDrivers now has the TrueSight.sys 2.0.2 sample, the old RogueKiller anti-rootkit driver @_CPResearch_ caught being abused at scale as an EDR killer. Load it, send IOCTL 0x22E044 with a PID, and the kernel terminates the process for you. https://t.co/LrEGF3u5Cu

byovdedr驱动工具 @0xocdsec 原文 ↗
更多96

制冷/医疗冷链成为攻击面,工控协议缺乏完整性校验的后果被具象化。

RT @MTSlive: SITUATION DETECTED: The EU refrigeration cyberattacks have escalated to medical infrastructure. Hackers breached Coolworld, a French national blood bank contractor, changing transfusion plasma deep-freeze temps from -30°C to 110°C. It was caught before the blood was boiled.

工控ot事件 @artem_i_baranov 原文 ↗

单 DLL 同时做到凭据截获与哈希诱捕,是可复用的 Windows 持久化思路。

I just created a basic DLL registered as a Windows Network Provider that captures all login credentials in plaintext. But that's just one capability, it can also inject fake servers into Explorer's Network view, and when the user clicks on them, Windows sends the NTLM hash to us. https://t.co/1gao6vbA90

凭据窃取ntlmwindowspoc @papadelospapas_ 原文 ↗

低成本硬件内置网络侦察、拦截与 AD 枚举,物理接入即用,值得试用。

ESPJack - a lightweight, command-line security toolkit designed for ESP32-S3 devices. Built for security research, penetration testing, and authorized network assessments. ✅ Network reconnaissance ✅ Traffic interception ✅ Active Directory enumeration ✅ Post-exploitation capabilities https://t.co/M3PxOjntbl

工具固件红队侦查 @ipurple 原文 ↗

供应链入口的实战回报被量化,附可参考的攻击链与赏金路径。

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

供应链赏金议题 @h4x0r_dz 原文 ↗

调试器滥用与检测工程素材,但内容封闭在限时邀请的社群内。

RT @ipurple: 🚨 Just shared the full playbook on abusing Console Debugger (cdb.exe) in the iPurple Discord. 🟣 If you care about Purple Teaming and Detection Engineering, we'd love to have you join the community. ⏰ Invite Duration: 24H ✉️ https://t.co/tQ1ppS2iAw https://t.co/Cl59RrUwan

工具检测windows @ipurple 原文 ↗

同一事件缺乏归因证据,只提供背景脉络而非可操作信息。

RT @grok: Unknown. No group has claimed the Coolworld/EFS plasma freezer attack, and French authorities (ANSSI, Health Ministry) have made no attribution. It fits a wider wave of unclaimed European refrigeration hacks on supermarkets. Some analysts flag possible Russian-linked actors based on scanning patterns and geopolitics, but others note weak evidence, potential false claims by groups like NoName057(16), or even opportunistic actors. Investigations continue.

工控归因 @artem_i_baranov 原文 ↗

常见混淆工具例行更新,仅在分析被混淆样本时相关。

javascript-obfuscator v5.10.0 — Obfuscates JavaScript and Node.js code with variable renaming, string encryption, control flow flattening, and anti-debugging to protect source code from reverse engineering. https://t.co/N0ju4HDHpd https://t.co/4N53uQ5bEA

工具混淆javascript @KitPloit 原文 ↗

把检测规则生成与编排交给模型的现成实现,但宣传性质浓。

Detection intelligence turbocharged with Al - Enable line-speed detection of cyberattacks by equipping your team with AI trained on 11 years of Detection Intelligence. Elevate your daily work from SIEM min-maxing to Detection Orchestration across Data Pipelines, AIDR, EDR, Data Lake, and SIEM. https://t.co/OHKYlRkdXt

ai_agent检测工具 @ipurple 原文 ↗

恶意样本语料分类口径变化,会影响依赖该仪表盘的统计结论。

I have finished training the new zrax classifier model and will reclassify all donated binaries over the next few hours; this may materially change ISA frequencies displayed in the dashboard. The new model is live on https://t.co/ndq85f45zf as well.

恶意样本工具 @in19h 原文 ↗

恢复能力扩展了远程修复面,对应急响应流程略有影响。

RT @WindowsUpdate: Need a fresh start for your Windows 11 PC? With Cloud rebuild, you can download and reinstall Windows and the drivers your PC needs directly from Windows Update, even if Windows can’t start. No USB drive or installation media needed. Learn more: https://t.co/B7RbTPSEO4 https://t.co/Ap6U0mseyU

windows取证 @artem_i_baranov 原文 ↗

沙箱逃逸是浏览器利用链关键一环,需按版本核对已修复。

RT @UpwindMDR: 🚨Critical - Apple WebContent Sandbox Escapes CVE-2026-112690 (App Store) and CVE-2026-113897 (WebKit Process Model) may allow remote attackers to escape the WebContent sandbox. 🔧 Apple’s Oct 26 fixes: improved input validation for 112690; removal of vulnerable code for 113897. 👉Fixed: • 112690: iOS/iPadOS 26.7.2 or 27.2; macOS Sequoia 15.8.2, Tahoe 26.7.2, Golden Gate 27.2; visionOS 27.2. • 113897: Safari 27.2; iOS/iPadOS 27.2; macOS Golden Gate 27.2; visionOS 27.2.

浏览器逃逸cve @0xocdsec 原文 ↗

生产环境未验证但条件极低,是域内横向的新切入点。

RT @unsigned_sh0rt: Quick blog on how iSCSI initiators/targets work in AD domains. If you can guess the right IQN you've got access to the disk. Haven't done the attack in prod but see the SPN shared with things like Hyper-V, Failover Clusters, and DFS.

ad横向移动工具 @0xocdsec 原文 ↗

implant 多样化的具体工程取舍,对检测侧建模有参考价值。

in addition, one of my theories for blackmagick-c2 is to increase the number of options per implant build. by increasing the number of different possible combinations of options, i gain resilience against signatures. thats why implant builds in bm c2 allow the operator to select which postex modules to include in that build (coff, bof, ffi, vm, and more). So you can include or exclude individual components, per build, and you have lots of different options for accomplishing the same tasks. The point is to make it harder to fingerprint any one build as that c2 or malicious.

c2免杀工具 @Teach2Breach 原文 ↗

栈清理归 caller 还是 callee 判断错,是逆向调试常见卡点。

🧠 Hook crashing the game? Stack corruption happens when your calling convention doesn't match. Knowing whether the caller or callee cleans the stack fixes it every time. More practical reversing tips from Guided Hacking: 👉 https://t.co/JsPMDj3iCI https://t.co/Bkw3y7AksZ

逆向调试 @GuidedHacking 原文 ↗

点明检测侧的不变量,对免杀与检测工程两边都有参考价值。

hooks are avoidable. kernel events not so much. but lots of programs do things like read files, so a single event is not an IoC. i’m not going to explain all the theory and tactics of evasion in this thread. whole books are written on it. but in general, you are never going to completely avoid leaving events and telemetry. you cant get the effects of calling an API without calling the API (in most cases), thats why the API exists. Yes we are limited to using the operating system like other desktop applications. The idea is to avoid signatured patterns, IOCs, and keep your “score” below a threshold where it gets bubbled up into an alert that gets investigated. The goal isnt to cause no kernel events. The goal is to blend in. The ultimate goal is to look like a normal process or application. The vm postex is likely to help with that. Its also likely to aid in resilience against defenders making easy signatures of your implants builtin postex commands. It is not silent. It is also just another option or tool in your toolbox. Its not perfect or necessarily better than other options in all scenarios. Its worth having as an option. In every scenario, details will vary and the best option is based on the environment you are operating in.

绕过edr议题 @Teach2Breach 原文 ↗

把脱壳与沙箱事件映射到反汇编视图,逆向流程能直接省一步。

RT @HexRaysSA: 🆕 Meet the Malware Analysis add-on for IDA Pro. Detects packed samples on load and offers to unpack them, emulates Windows code safely with no VM, and maps sandbox events onto your disassembly. Grab it for 50% off with promo code MALADD50 *See comments for rules and restrictions. 👉🏽 Check out the feature set and case studies here: https://t.co/zBeqzYnznz

工具逆向沙箱 @mrexodia 原文 ↗

权重级定向篡改可改变模型行为,是模型完整性的现实威胁样本。

RT @HaozeHe_Hector: We edited 0.0002% of NanoJev’s weights, and the edited model beat the original 65–34 (1 draw) in 100 head-to-head Tank War games. 🎯 The original took only about 1 in 3 clean shots. Using Weight-Space Programming by @rikiyatakehi et al. (https://t.co/oQW1HI5rU9), we read the fire-vs-hold direction out of its activations and added it to the decision head: 1,024 weights, no training. Now it takes every clean shot and fires about twice as often. It isn’t a better shot, though. It just stopped hesitating. 🛠️ Weight editing is now a trainer in Honing. An edit goes through the same gate as SFT and RL, and only ships if it beats the current model. Report: https://t.co/xpLcLMauje Weight-edit trainer: https://t.co/hoXdxumA1s GitHub: https://t.co/oAPAOkCpcy Follow @HoningLab for more Honing updates :)

llm模型篡改研究 @morluto 原文 ↗

是 Agent 技能分发的实例,与安全无直接关系。

RT @Miguel07Code: I loved @morluto REA and we now open-sourced RAV, /reverse-any-video with @HyperFrames_! update the skills with npx hyperframes update reverse-any-video and use it! some of crazy good examples here with a Stripe launch video fully 1-shot https://t.co/10K4hdhtfM

ai_agent工具 @morluto 原文 ↗

macOS 漏洞赏金生态的进展,无具体漏洞细节。

Congrats on the milestone! 🥳 And mahalo David (@dps) for reaching out & being so open to candid bug bounty feedback! 🙏🏽 Stoked our Foundation's work led to more ack'd bugs. Payouts TBD; but w/ @Meta's 2x match for non-profits hopefully turning 🐛 into 💰 for @objective_see! 😇

漏洞赏金macos @patrickwardle 原文 ↗

macOS 可实时拦截进程与持久化,iOS 该能力缺位是结构性防御盲区。

RT @ojstudioapp: @patrickwardle @AppleSupport @Apple Endpoint Security is what lets Mac tools like BlockBlock watch process launches and persistence attempts in real time. iOS has no equivalent API, so defenders are mostly left with after-the-fact diagnostics. That Lockdown Mode thread is a good example of why the gap matters.

移动端绕过议题 @patrickwardle 原文 ↗

用的是 n-day 而非 0-day,补丁滞后本身就是可被利用的入口。

Darksword just never seem to die. People have paid for it, now they have to use it !? I really wonder how effective those exploit chains still are ...but I guess, with all the vibe coded C2 frameworks out there, it is a low hanging fruit approach: https://t.co/ug5apy1bMS

移动端cve报告供应链 @NetAskari 原文 ↗

把 LLM 恶意软件分析结果做成可复核证据,降低幻觉误判。

RT @SentinelOne: AI can reverse engineer malware. We'd still like to see the receipts. @LabsSentinel developed VerifiedIDA to keep LLM-driven analysis tied to evidence, open to review, and reusable across security workflows. Explore the full research: https://t.co/zPiqAPtcaE

llm工具报告 @_winterknife_ 原文 ↗

把崩溃日志转成可利用性判断,iOS 漏洞研究门槛较低的切入口。

RT @8kSec: Learn how to read and analyze iOS kernel panic logs, extracting them, reading the panic string, calculating real addresses from the kernelcache slide, and symbolicating the backtrace: https://t.co/UUcdfLFgzn  The example panic ties to a PUAF primitive, showing how one crash log can point you to an exploitation-related condition. Follow @8kSec for more iOS kernel research

ios内核报告 @0xocdsec 原文 ↗

完整记录漏洞利用可靠性调优工程,对稳定性问题有借鉴价值。

RT @Grimdoomer: Here's part 3 of my Hacking the Xbox 360 Hypervisor blog series. This part covers all the improvements made to the Bad Update exploit to get it near 100% reliability. Part 4 which covers the Peer Pressure exploit will be out in the next week or so... https://t.co/tXfgb8T9U7

内核漏洞利用报告 @alkalinesec 原文 ↗

AI 红队评估能力被大厂收编,反映 agent 安全测试的资源流向。

RT @jonsakoda: We are excited to announce that @CrowdStrike has signed a definitive agreement to acquire @dreadnode, the leader in cybersecurity evaluations for frontier AI The company will join CrowdStrike’s Cyber Superintelligence Lab to combine AI security research with autonomous cyber defense for the first time 🔥 Congrats to Brad Palm (@bradpalmtree), Will Pearce (@moo_hax), Nick Landers (@monoxgas) and the team!

ai_agentllm事件 @dreadnode 原文 ↗

AI 品牌仿冒已规模化,商店审核对这类滥用几乎没有防护。

RT @privacyis1st: 📓 “NotebookLM? Absolutely not. We moved the arch three pixels. Totally original.™” 😂 ⚠️ FIVE Google NotebookLM lookalikes made it through Apple’s App Review team. Five separate publishers. Familiar blue/purple arch icons. Notebook/LM names. Recurring subscriptions. And the confusion already has a reported financial consequence: one reviewer says they paid believing the app was Google NotebookLM. The developer’s public reply acknowledges the confusion and explains that it’s a separate product. Another reviewer accuses a second app of impersonating Google’s service. 💸 Weekly subscriptions range from $4.99 to $9.99—up to approximately $521 a year at the weekly rate. 🚩 The investigation also found a “Terms & Conditions” page serving privacy text, a policy naming a different app, an unfinished governing-law clause and a broken support link. This is a suspected-scam warning backed by original evidence. Five similar icons do not prove one operator—but they do show a recurring problem that users can encounter while searching for Google’s product. 🍎 Apparently, originality now comes in five slightly different shades of blue. The five listings: 🔗 AI Notebook: Quiz & Flashcards - https://t.co/KIFww9vFMI 🔗 App for AI NotebookLM & Quiz - https://t.co/bj5sKfPMIb 🔗 Notebook: AI Quiz & Notes - https://t.co/ukmWTiDUY5 🔗 Notebook Flashcards & Quiz LM - https://t.co/AB5UcxMUp1 🔗 Notebook Quiz & Flashcards LM - https://t.co/QYhU0dsZql 🔎 Full investigation, icon comparisons and review screenshots - https://t.co/azKNmK7Gl4 Hey @Google @Apple @AppStore — five familiar-looking icons, five separate sellers. How many arches does it take to trigger an App Review? 📓 #AppStore #Apple #NotebookLM #ScamAwareness

移动端仿冒llm @patrickwardle 原文 ↗

xen — Open-source type-1 hypervisor and Virtual Machine Monitor providing secure isolation for x86, ARM, and RISC-V, with Dom0/DomU architecture, xenstore, and Intel TXT trusted boot support. https://t.co/uyeeyN2r27 https://t.co/9D7DQziXmV

虚拟化工具 @KitPloit 原文 ↗

I once got the role of being a red teamer for a very large genealogical company for several years due to the previous team accidentally hacking a 3rd party payment processor during their engagement. I asked them: what happened to the previous team? They answered: we called law enforcement, they were arrested. We figured, if they were that bad at their work, the last thing we would ever want is them to continue being terrible at their job at the expense of others.

红队事件 @Laughing_Mantis 原文 ↗

Another toy project. A cross-platform tool for managing PATH and environment variables on Windows and Linux. It scans your environment for broken or duplicate paths, shows when different versions of tools like python or node are shadowing each other, and lets you reorder or edit everything in one place. Every change is backed up with a snapshot so you can roll back anytime if something breaks.

工具环境变量 @kernullist 原文 ↗

RT @pavandavuluri: Today, we’re sharing updates from the team on memory improvements, including work that makes experiences on 8 GB @Windows 11 PCs faster and more responsive. Windows quality is an ongoing journey. We’ll keep listening, building on what we learn, and sharing our progress 🚀. https://t.co/4CQOSspoaS

windows产品更新 @artem_i_baranov 原文 ↗

搜索常见软件即可能中招,安装器级持久化与变现手法有通用参考价值。

Microsoft Threat Intelligence has identified an evolving fake-download campaign using trojanized Inno Setup-style installers and targeting users searching for cracked commercial software, free games, activators, plugins, and unofficial utilities. https://t.co/LJVInHDi4k https://t.co/GWdXmg6gyr

恶意软件供应链报告 @virusbtn 原文 ↗

全程只使用 Windows 内置工具的完整入侵链,可直接对照做检测规则。

Bitdefender's Marius Baciu, Adrian Schipor & Martin Zugec show how a 3AM (ThreeAM) hybrid AI ransomware affiliate breached a European school estate in the UK, then used ordinary Windows tools to build domain trusts and reach other schools. https://t.co/d7igKf5Zsl https://t.co/PJztytCD9o

勒索软件横向移动报告 @virusbtn 原文 ↗

借系统自动填充隐藏订阅价格、跳过用户确认,已在四国出现。

RT @GenThreatLabs: The scam page can hide the price. The SMS can’t. Unless your phone does it for them, the code is offered for autofill before you ever open the message that says “every week”. We call it #BillBait. Seen in 🇵🇱 Poland, 🇨🇭 Switzerland, 🇩🇰 Denmark and 🇨🇦 Canada. The mechanism is one line. Beyond the regulated customer text, the message carries the domain and code in a machine-readable pair: "@​example[.]com #2976" iOS and Chrome match it to the site you’re on and put the code right above the keyboard. No opening a message, no retyping. You tap the suggestion — and you’re subscribed. The SMS stating price, frequency and unsubscribe instructions never got read. For a login code that’s convenience. For operator billing that code IS the payment authorization. No card, no 3DS, no second confirmation. It starts a recurring charge on your phone bill. iPhone showed 14.99 zł and dropped the “every week”. Chrome showed everything inside a prompt wearing Google’s authority and designed to be dismissed in half a second with an approval. How people get the message in the first place? Ads with large font saying “Verify your phone to get full access”, 14.99 zł every week in tiny text under the banner. Not an accident. Open the whole SMS. Check the frequency. If it’s not what you clicked for, don’t submit. And check your bill. An unfamiliar premium SMS or operator billing line means something got authorized. Call your operator, ask who’s collecting, cancel it, dispute the charge. Marking the sender as spam doesn’t stop the charges. Full analysis: https://t.co/MPmk23JU8v #SMS #MobilePayments #DCB

钓鱼移动端浏览器诈骗 @virusbtn 原文 ↗

正规 RMM 工具被武器化,白名单软件滥用检测思路可迁移。

Cato CTRL researchers investigate a suspected MuddyWater campaign targeting Israeli organizations via phishing emails impersonating the Israel Tax Authority. The campaign abuses SimpleHelp, a legitimate RMM tool, to enable attacker-directed remote access. https://t.co/hBCbONgAwf https://t.co/6M7yX57vXm

钓鱼rmmapt @virusbtn 原文 ↗

低复杂度工具链依然奏效,说明基础投递与反连检测仍是薄弱环节。

RT @ESETresearch: #ESETresearch has identified #Gorbag, a Russia-aligned 🇷🇺 threat actor that uses simple infostealers and a reverse shell to harvest data from Ukrainian 🇺🇦 organizations near the front line: https://t.co/HUizwkRobw 1/6 https://t.co/EqG82runOD

apt窃密报告 @virusbtn 原文 ↗

可直接落地的 IOC 清单,含 Cobalt Strike、AdaptixC2、vshell 等标注。

RT @teamcymru_S2: INTEL DROP Malware-hosting IPs in the Total Insights detection pipeline. 12 hosts that also run live C2, offensive tooling, or phishing. Spread across China, Vietnam, the US, Japan, Nigeria, and Indonesia. Every IP below is tagged malware-hosting plus the tags shown. 1.92.135.168 cobaltstrike + adaptixc2 + socks + ai-relay:new-api 101.200.193.211 vshell + cobaltstrike + proxy:nps 106.75.7.239 cobaltstrike + vshell + java-chains + socks 103.27.239.184 quasar 107.175.39.140 mirai + open-dir 134.122.140.63 cyberstrikeai + open-dir 111.228.5.127 cobaltstrike 110.42.215.163 artex + open-dir 102.220.161.143 phishing 113.45.207.71 iot:ip-camera + socks 104.238.147.251 proxy:frp + iot:home-assistant 103.172.186.175 scanner + router:zte #TotalInsights #ThreatIntel #MalwareHosting https://t.co/nDqrsELahM

威胁情报c2ioc @virusbtn 原文 ↗

边缘设备预认证 RCE 的收购热度,预示该攻击面价值上升。

RT @SecuriTeam_SSD: F5 BIG-IP Pre-Auth RCE is now in high demand! Found a vulnerability of this kind? Now is the perfect time to submit your research. Check out our full list of high-demand targets and see what we're actively looking for: https://t.co/gjeLmHkw5E

rce漏洞市场网络设备 @0xocdsec 原文 ↗

RT @MsftSecIntel: Microsoft Security Research has identified a fake-download campaign targeting users searching for cracked software, free games, plugins, and other utilities to distribute trojanized installers that can monetize compromised devices in three different ways: https://t.co/d84ryG3avP - Routing third-party internet traffic through the compromised devices - Enrolling devices in a distributed-computing platform - Installing additional utility software The trojanized Inno Setup-style installers launch executable payloads staged in temporary directories and track each component’s installation outcome separately. A failed or blocked installation does not necessarily stop the rest of the bundle. Organizations can reduce exposure to this activity by blocking the use of untrusted software sources, enabling potentially unwanted application (PUA) protection, and monitoring for unexpected software installations and persistence mechanisms. Read the blog for technical details, indicators of compromise (IOCs), mitigation guidance, Microsoft Defender detection info, advanced hunting queries.

恶意软件供应链 @virusbtn 原文 ↗

把苹果近年加固机制的补丁差异与信任边界整理成可对照的地图,不含利用。

RT @azozzalfiras: 🔬 Just published my iOS security research repo. Static RE notes, patch-diffs & trust-boundary analysis on modern iOS: • SPTM / TXM monitors• Secure Kernel exclaves • Kernel zone mitigations • Media parsers (ImageIO etc.) No exploits. Just the map of how Apple hardens things. Feedback from RE / kernel researchers welcome. https://t.co/zDzp5MnVge #iOSSecurity #AppleSecurity #ReverseEngineering #iOS #SecurityResearch

移动端逆向报告 @0xocdsec 原文 ↗

一线数据点:让 Agent 复现人工发现的高隐蔽漏洞仍然吃力。

We still have a couple guys that find more bugs without using AI than they can find when they do use AI And the bugs they find by hand also seem to be, in most cases, quite illusive when one tries to replicate the search with AI; even when one really leads the agent to water

ai_agentllm报告 @C2IRIS 原文 ↗

可跟着复现的 Windows API 实战代码,红蓝队常引用该系列做对照。

https://t.co/BNXRZjsvEY next one from my blog! #windows #malware #malwaredevelopment series! enjoy! ❤️ https://t.co/lLNYQmlgLN #hacking #research #purpleteam #winapi #programming #malwareanalysis #threatintel #redteam #blueteam #purpleteam #ethicalhacking https://t.co/nnq9whyyju

windows工具报告 @cocomelonckz 原文 ↗

签名密钥泄露波及 FRP 信任链,绕过思路与工具配置已公开。

RT @AlexAltea: eMudhra key leaks impact Android's Factory Reset Protection (FRP), thanks to the generous GMS sign-in interface. TL;DR: Inject mm.setContentFilters('{"user_setup_complete":1, "device_provisioned":1}'); on server JS responses with `mitmproxy --emudhra` and reboot, bypassing SUW. Then, flip system_server's mFrpActive with any N-day to leak the secret and wipe FRP. Demo below. Full chain private for obvious reasons.

移动端绕过密钥泄露android @0xocdsec 原文 ↗

教室物联网设备被远程植入监控模块,暴露固件与远程下发信任问题。

RT @nexta_tv: 👀 Secret surveillance software discovered in Moscow schools: interactive whiteboards can record students and teachers A Moscow teacher discovered software installed on interactive classroom whiteboards that can secretly activate the microphone, record conversations during lessons, and send audio files to servers operated by the city's Department of Information Technology. The software has reportedly been installed remotely since August 2026. Every ten minutes, it checks for commands from a server. After recording, it deletes the audio file from the device. Neither students nor teachers receive any notification. Here's the interesting part: the software was originally called “Lesson Audio Recording Agent.” It was later renamed to the much more innocent-sounding “Lesson Support Agent.” According to Mediazona, traces of the software were found in at least three schools. A source told the outlet that the recordings were intended to be analyzed using AI to monitor what was being taught in classrooms. Moscow's Department of Information Technology did not respond to journalists' questions. The scale of the system's deployment and what happens to the recordings remain unknown. In Russia, even a classroom whiteboard can now become an informant.

固件监控物联网事件 @0xocdsec 原文 ↗