Must-see26
微软未修复的 Windows 本地提权,PoC 已存在且补丁被再次绕过
RT @karollooool: MSRC dismissed my LPE because it supposedly needs dev mode. Funny because I already sent them an updated working PoC without it like a month before they dismissed it.
I appealed on Oct 15. but that obviously went into the void…
There is still one small detail I’m not 100% sure about, but they didn’t even mention that in their dismissal either.
And now today I managed to bypass what looks to me like a deliberate fix they introduced in build 29695.1001 (Oct 23) so it comes down to SYSTEM again…
Like seriously, what the actual fuck are you trying to do here @msftsecresponse ?
If anyone could connect me directly with someone from their team I would really appreciate it, because I just wanna get this sorted out already somehow. DMs are open.
实证 LLM 自动化密钥猎取的产出效率,暴露密钥治理缺口
RT @deedydas: An abliterated LLM just found me ~$3,300 in live ElevenLabs keys on the internet.
All the OpenAI / Anthropic keys it found were already drained which makes me think they've already been actively monitoring this. Pretty insane. https://t.co/HcNhxudDqm
本地可自托管的无审查安全研究模型,降低红队 LLM 工作流的部署门槛
RT @zroai_: We’ve released the weights for Dolly 2 Security, our uncensored GLM-5.3 model built for cybersecurity research.
We took a targeted approach to abliteration, aiming to reduce refusals while preserving the model’s general capabilities. Download the FP8 weights and run them yourself, or use dolly2-security on Zro without setting up GPUs.
https://t.co/mFP19DXuLo
LLM 自主挖洞已产出可直接 root 服务器的完整链路
RT @aegiryy: this blog shows some of the worst bugs claude found, including a zero-click rce on the linux networking stack that roots a server over the network with no creds or interaction, and tens of full rce chains in chrome (recording below is the one in the blog): https://t.co/DHkfR22s94 https://t.co/J9noyTTxZI
原厂补丁未落到设备,暴露面横跨 OT 与边界设备
RT @4A4133: Announcing 3 Critical vulns for hundreds of millions of IoT, Firewalls, OT, and Military devices including: Palo Alto Networks, Comcast, D-Link, HP, Canon, Honeywell, GE, Oracle, Siemens, Rockwell, Digi, Lockheed Martin, Boeing...
The vulns are with GoAhead Web Server which is a backend library used by these devices. There's a patch available from GoAhead but that doesn't mean there's a patch available for the devices yet -- that's up to the manufacturer.
- TLS traffic over DHE ciphers have a 1-byte secret, "04", meaning anyone can decrypt connections to these devices and steal login credentials and session tokens if the client negotiated with DHE ciphers like the latest TLS_DHE_RSA_WITH_CHACHA20_POLY1305_SHA256, which is extremely likely.
- LibreSSL mTLS has been completely bypassed, meaning anyone can connect over mTLS. If you have one of these devices listening on the internet and are expecting mTLS to protect it from attackers, take it offline or patch asap.
- OpenSSL Certificate Revocation Lists (CRLs) are not enforced, meaning if a user device is stolen or compromised, and you revoke that mTLS cert, nothing actually happens. There are a lot of other attacks this can be used for.
These are critical because these types of devices run our world. Water systems, electricity, military, firewalls, critical infrastructure.
https://t.co/AYUgNsrA0f
位于启动链最底层,受影响面覆盖主流服务器固件
RT @arstechnica: Code-execution flaw in UEFI firmware threatens most servers on planet Earth https://t.co/wdjLscqSY2
僵尸网络已把新 CVE 写进现成利用routine,自托管 Linux 服务处于暴露窗口。
Interesting finding from my colleagues for anyone following #PHP2SHELL activity
They found a Linux botnet sample that already has code targeting CVE-2026-109618 integrated into its existing exploit routines, alongside eight other CVEs
给出完整 RCA 与触发路径,可据此判断同类解析逻辑是否受影响。
RT @immortalp0ny: CVE-2026-50494 ntfs!AnalysisPass RCA https://t.co/cZDnuGcIFz
披露节奏与协同模式变化,直接影响自托管组件的补丁响应窗口。
RT @k8em0: PHP2Shell had everyone buzzing last week, but the multiparty vulnerability disclosure felt all too familiar. Dan Kaminsky's DNS bug was the first attempt at such large scale coordination and we should expect more like this from AI-assisted research
https://t.co/aDpdQO3gb2
可直接接入 SIEM/EDR 做威胁狩猎,省去自建扫描管线。
C2IntelFeeds — Automatically generated C2 threat intelligence feeds of IPs, domains, URL paths, and implant configuration metadata, derived from internet scanning for SIEM, EDR, and threat hunting. https://t.co/kFZL9508hE https://t.co/dldMCgmmTa
未登录即可从 SQLi 打到 RCE,蜜罐已见探测,十月 CPU 外仍有未修补实例。
🚨 Our Oracle E-Business Suite decoys recorded exploratory exploits against CVE-2026-101767, an unauthenticated SQL injection in EBS's Thin Client Framework servlet that chains to remote code execution, this morning, Oct 27.
The bug was recently disclosed by @Horizon3ai and patched in Oracle's October CPU. No public report exists of this CVE being attempted in the wild.
Full IOCs on Defused Radar 👉 https://t.co/TTnxgi9Hv5
签名有效的 RMM 滥用可绕过应用白名单,IOC 与落地域名已公开。
EV signature "Space Solar Technologies LLC"
Signs fake CloakGuard RMM disguised as Windows service -> fake blue screen -> installs ScreenConnect
rmm[.]cloakguard[.]org
162[.]246.19.216
relay[.]alico[.]shop · 69[.]169.106.96:8041
https://t.co/ozQGmJR3Gi
可自托管的 AI Agent 自动读源码并实打 exploit,是 Agent 攻击面的现成样本。
shannon v3.5.0 — Shannon is an autonomous, white-box AI pentester for web applications and APIs. It analyzes your source code, identifies attack vectors, and executes real exploits to prove vulnerabilities before they reach production. https://t.co/7jDBm6BuQ9 https://t.co/7EtCv6GEOA
没有攻破 Anthropic 基础设施,仅靠用户信任 AI 域名并粘贴命令即可投递
RT @GenThreatLabs: Two malware campaigns are abusing Claude's content-sharing features and paid search ads to distribute malware from Anthropic's own domain. Neither campaign compromised Anthropic infrastructure. Both rely on users pasting a command they did not write.
The pages are published as Artifacts and Shared Chats on https://t.co/a0p0AUOKxG and pushed to the top of search results with Google Ads.
Delivery uses #ClickFix: victims are instructed to copy an obfuscated command and run it themselves. On macOS this chain installs Atomic macOS Stealer (#AMOS), which harvests browser credentials, Keychain data, crypto wallets and other personal data, then replaces the Ledger and Trezor desktop apps with trojanized versions that ask for the recovery phrase. On Windows the same pattern delivers a
.NET remote-access tool that also targets cryptocurrency wallets.
No exploit was required. The attack depends entirely on user action, which is why the most effective control is: do not paste commands from a browser into a shell.
Full analysis: https://t.co/ezNlaAHlsn
ai_agentllm钓鱼事件
@artem_i_baranov
source ↗
把 AI 工具名当作信任凭据,DGA 加 DoH 取载荷的链路完整可复现
Tricks from the past via signed ChatGPT and Claude binaries.
🎣Fake ChatGPT and Claude projects on GitHub
📦Signed binaries sideload Yoba malware
🖼️Telegram channel image seeds a DGA
🔐DNS over HTTPS retrieves payload instructions
Full research 👇:
https://t.co/qeabCNSUzM
原生插件在运行时解密,并在 CI 与沙箱中休眠,常规检测很难发现
RT @JFrogSecurity: 🚨 testmetest goes beyond npm.
JFrog Security Research identified 42 armed GitHub repositories across 11 legitimate owners hiding an npm publisher, plus a new native addon that decrypts its payload at runtime and stays dormant in CI, sandboxes, and analysis tools.
The campaign spans 25 npm package names. Check whether you installed any testmetest or @iobandamwer packages and remove them if found.
Full analysis: https://t.co/oLtvs4L3wU
干净的 2.4.7~2.4.9 环境均可复现,无需认证即可拿下电商站点
RT @sansecio: 🚨 We discovered StyleSmuggler, an unauthenticated RCE in Magento & Adobe Commerce under active exploitation. Confirmed on clean 2.4.7, 2.4.8 and 2.4.9 installs. PHP executes during failed payment email rendering.
Our research: https://t.co/8hbsdP56K9
公开分步给出触发、信息泄漏与写原语,是可直接跟做的内核 LPE 教程。
Linux Kernel UaF to Root
Part 1: https://t.co/iWVoizBw7p
Part 2: https://t.co/KbPDa6Na7X
Part 3: https://t.co/c9qv8URcn8
Credits Quentin Buathier
#infosec https://t.co/7a9Pqn7IfV
域名可直接转成 DNS 阻断清单,覆盖加国主流银行的 2FA 仿冒。
👇🏦🎣
td2fasecure[.]com
cibc2fasecure[.]com
rbc2fasecure[.]com
bmosecure2fa[.]com
避免按已被证伪的归因做应急,相关 CVE 遭利用的说法目前证据不足。
RT @Unit42_Intel: 2/2 This led us to believe that it was exploitation of those vulnerabilities. We are unable to confirm that initial access is associated with the specific PHP2Shell CVEs. We have retracted our previous update from October 23.
AI 驱动逆向直接接进主流反编译器,Agent 化 RE 工作流已落地
RT @HexRaysSA: 📢 IDA 9.5 is here: more platforms, more agentic power.
→ 3 new decompilers: Android Dalvik, Infineon TriCore and Qualcomm Hexagon
→ IDA MCP and Assist for AI-driven reverse engineering
→ Recursive decompilation across call graphs
→ IDA Home now includes 11 cloud decompilers instead of 2, at the same price
→ And much, much more
It’s one of our biggest releases yet. Active license holders can download it now from the Download Center.
https://t.co/sNYObPyqDz
移动端 hypervisor 远程利用链公开,属基带之外的新攻击面
RT @offensive_con: Netflix and Pwn, Remotely Exploiting MediaTek Phones via the GenieZone Hypervisor by Philipp Mao and @s_k_u_u_k https://t.co/JhPV9dAGW3
私钥可穷举,任意站点 HTTPS 证书都能伪造;受影响设备仍在广泛部署。
RT @IntCyberDigest: ‼️ Cisco says its IOS XR routers and Meraki cameras trust eight eMudhra certificate authorities whose private keys were numbers below 256.
Anyone can recover those keys in seconds and forge HTTPS certificates for any website.
Cisco is still investigating its ASA firewalls. https://t.co/Iij87zHWPX
CoreGraphics 零日存在实际利用迹象,iOS/iPadOS 需升至 18.7.11。
RT @ApplSec: 📣 EMERGENCY UPDATE 📣
Apple pushed additional updates for a zero-day that may have been actively exploited.
🐛 CVE-2026-86950 (CoreGraphics) additional patches:
- iOS and iPadOS 18.7.11
签名与模板式 EDR 检测在运行时生成代码面前失效,攻防重心正在移动
RT @Teach2Breach: wrote about this years ago, but runtime code generation is the next thing. everybody scoffed when i first suggested human language hacking with ai controlling the c2, now there are c2s with mcp designed for it. the next logical step is postex code generated at runtime based on the conditions of the compromised environment. EDR based on templates and signatures will fail
模型推理链提取已被明码标价,自建 LLM 服务的推理泄漏面有了价格参照
RT @lyq_sqsp: 6.6K bounty 😅 afaik someone is willing to pay >6.6M for a large scale reasoning extraction from Opus 5.5
More79
钓鱼投递链条中绕过默认拦截的可行手法,附带完整复现步骤
Last week, I wrote about Custom Extensions and how they can be used to bypass SmartScreen protections.
Curious how this technique works? Read the article below 👇
https://t.co/9Q7f7Dzhby https://t.co/3fqT3NlUMn
导致自动化证书吊销检查静默失效,影响依赖 CRL 的客户端与流水线
RT @kingslyj: eMudhra has been blocking curl/wget/python urllib User-Agent headers from accessing CRLs
https://t.co/C48eRbreLE https://t.co/Yh6wqhFFW0
路由基础设施暴露面的可量化研究,附完整 PDF 数据
RT @Enno_Insinuator: Detecting and Characterizing Exposed BGP Routers
https://t.co/q2lolJseNW [PDF] https://t.co/eidNKQvRic
移动端沙箱与 JIT 限制的突破口,涉及 iOS 安全模型边界
RT @khanhduytran0: iOS 26.0 simulator runtime running natively on iPhone, sandboxed, works without JIT
It starts from launchd_sim then spawn other real simulator processes like backboardd, SpringBoard, etc https://t.co/j6QTKTmz3s
若属实则端侧本地推理选型将变化,影响自托管 Agent 的模型部署
RT @noctus91: bro how is a 1.7B model beating models 2-5x its size across phones and laptops 😭.
at this point Liquid AI is just bullying Gemma and Qwen . https://t.co/eHu9ESUgIz
云厂商权限审批流程的执行缺陷,但仅单方爆料无技术细节
RT @Conspiracookie: 🚨BREAKING (EXCLUSIVE): National Security Concerns Leaked by MSFT Whistleblower
Leaked Microsoft screenshots show an India-based engineer received approval to view U.S. government cloud logs.
A Microsoft whistleblower contacted me and shared screenshots supporting allegations that Microsoft used its “escort” process to allow an India based engineer to access Government Community Cloud (GCC) systems. The whistleblower alleges the engineer is not a U.S. citizen.
Microsoft’s own published rules state that temporary elevated access to customer content in GCC requires verified U.S. citizenship. The escort model is the workaround the company uses so non-U.S. personnel can still direct the work.
The screenshots show Viraj Chokhany requesting and receiving approval for escort access to view logs in the DPRODMGD151 environment, which is part of Government Community Cloud Moderate — the cloud used by many federal, state, and local agencies.
In the escort process, a U.S. person is supposed to open the system while the India-based engineer guides the work.
The whistleblower believes this raises urgent national security concerns and may violate FedRAMP requirements.
Multiple Microsoft employees have reviewed the screenshots and confirmed they appear authentic. Viraj, has been confirmed to live in India-- Where he also attended and graduated college.
The whistleblower is working on a public statement. I will release the whistleblowers statement when I receive it.
Attached images are some of the screenshots that I received from the Whistleblower.
@DHSgov @FBIDirectorKash @JDVance
观点性论述,无新数据,仅作行业反思参考
One hill I'll happily die on: the normalization of "ship fast" culture at the expense of proper SDLC has played a significant role in the quality of vulnerable code being pushed into production and the rise of CVEs and enshitification of software for the last several years.
We've systematically traded engineering rigor for velocity, security assurance for investor confidence, and long-term maintainability for quarterly growth. All things I wish we would second guess and reevaluate.
But the software overlords wanted products yesterday and they got them. And now the rest of us get to spend the next several years discovering exactly which corners were cut and figuring out how to deal with the immense amount of responsibility that's been dropped on our collective feet.
反映不同模型在硬件改装请求上的拒答差异,细节不足
RT @0xSero: Such a shame Opus refused to unlock my RTX 3090, there was an unlock found a few days ago...
Deepseek won't say no, though. https://t.co/OO5qfZo3uo
一句话防御侧提醒,无技术细节
RT @j2k3k: you don't need daybreak, you need to segment your fucking network
情绪化行业观察,无实质信息
RT @techspence: I feel bad for this person. We need sysadmins now MORE THAN EVER. They are like the last adults left in a room full of crazed AI lunatics... https://t.co/L1dOf0uh2U
无证据的阴谋论式说法,仅作舆情感知
RT @HackForumsNet: @tonedeafandh0t They are monitoring everything on your phone.
Don't you get it?
Apple and Google phones are complete surveillance devices. Cameras, texts, and audio are all being constantly monitored and recorded.
无上下文短句,信息量近零
RT @social_sth: Deutsche Ingenieure haben mal wieder gekocht!
云原生重定向可绕过基于域名与信誉的检测
RT @ipurple: Azure API Management as a Policy-Based C2 Redirector https://t.co/oYBSlAUkm5
内存态账户隐藏手法可直接用于驻留与对抗排查
RT @init1security: Windows account hiding 101: A $ suffix skips net user but fails against PowerShells: Get-LocalUser.
To fully automate deeper registry hiding, @wgpsec dropped this: https://t.co/JnjTke7qGP
To avoid standard Windows disk artifacts, we translated their logic into a BOF and C# implementation to stay strictly in-memory. Definitely adding this to our training stay tune!!🕵️♂️
macOS 逆向可据此切分代码边界,便于分段分析
cud recovers probable code units -- translation units, compilation units, object-file groups -- from Mach-O executables, dylibs, and bundles. https://t.co/s5ic2eC6c3
RT @shotgunner101: This is why I went GrapheneOS when I did. Besides the targeted ads I would get through my Samsung phone I was seeing the trend of phone makers and governments pushing for clientside scanning into phones and increasing levels of telemetry.
The fact your phone can decide your now allowed to see your partner in an intimate manner cause its doing clientside scanning of your private conversations should concern EVERYONE.
This same tech system is used in China to surveill citizens to catch them discussing the government in negative manners, to catch them sharing or videoing protest symbols/images, etc.
These companies will claim its protect people but in reality is to surveillance, censor and eventually control them. 💀
RT @vx_herm1t: Metamorphic compiler https://t.co/MDmFCQ80Do https://t.co/18hzCHq33a
RT @MSNightmare2000: @blahmoah unprivileged user convinces defender to create arbitrary files with convenient DACL for LPE, the next 0day
Wanted to remind everyone that prompts are no longer a level of effort or skill, since they are greatly being generated behind the scenes at frontier levels, as told from someone who works with them directly
https://t.co/PpapoAc1fv
@artem_i_baranov @OfficialLoganK needs to figure out daybreak blue or red like programs that work, its unclear how they even want to give their cyber models. there is fairwind but no one seems to get access
RT @shanejcaldwell: When working on agent control solutions, you should publish what the cost in capability was. If you don’t know, your customers will definitely let you know!
固件级 AES 实现的真实攻击路径,嵌入式与 IoT 选型可作参考。
A side-channel analysis of the Arm CryptoCell-310 AES (Guillaume Bethouart)
https://t.co/FY2aHJAwjY
#infosec https://t.co/RVDMbPgiiQ
完整加载链拆解,可作检测规则与样本分析的取材。
RT @InvokeReversing: We've uploaded our live stream from October 16th where we analyzed a Rust loader that uses indirect syscall resolution, AES decryption, ETW patching and multiple unpacking steps to load and execute NWHStealer. Enjoy! https://t.co/Ju5JqwAM81
零点击投递与广告分发渠道结合,是投放链条的新形态样本。
RT @jsrailton: UPDATE: Israeli company we caught delivering zero-click infections is #1 bidder on Starlink ads in 105 countries.
Data from @trendai_RSRCH suggests worldwide surveillance op targeting @Starlink users. 1/
By @razhael
https://t.co/uWFzemhgnL https://t.co/FxkDlK7Xo0
AI 攻击者假设下的防御范式讨论,影响 Agent 权限与审计设计取舍。
RT @PrototypeVC: The internet’s current security regime, which was built on the assumption that attackers are human, is obsolete in the face of AI. https://t.co/LBXHU3vAYc
"there are teams running models... without a custom sampler"
"AAHHH THATS SCARY!! STOP!!!" https://t.co/1SPR1wEL32
I see some people sharing on a weekly basis, evasions against popular EDRs stating that they did it due to White Knight Labs.
Is this a new marketing trend? 🤷
Gitea
https://t.co/ap7Wb8tkq4
libslirp 影响 QEMU 与容器网络栈,文附可复用的 fuzzing 方法与开源代码。
RT @quarkslab: Fuzzing complex state machines with LibAFL-QEMU and incremental snapshots.
We walk through our approach, revisit CVE-2026-9539 in libslirp, and explain why our initial fuzzing campaign missed it.
Fuzzer code is open source 🧑💻
🔗 Check it out: https://t.co/gQCwuSy7Dv
红队隐蔽出网通道开源,防御方可据此提炼 S3 异常的检测特征。
An AWS S3 UDC2 egress channel for Cobalt Strike is now open source.
Our demo covers the AWS infrastructure, Terraform deployment, encrypted Beacon traffic, and dynamic pre-signed URL refresh.
Watch the demo: https://t.co/uzuEtxYLYl https://t.co/WhRxx4Qga7
PPL 加载原语已具备,缺的只是可用的签名 DLL,思路具备复用价值。
I can load any signed DLL into PPL protected processes.
However, I can’t exploit this because each DLL checks the service name and enters an idle loop on mismatch, without exposing RPC, named pipes, or useful functionality. It’s running as PPL but offers no surface. Any ideas?
可直接上手的逆向辅助功能样例,缩短熟悉 IDA 新能力的路径。
When presenting Hex-Rays IDA Assist internally to the team, I built a set of sample programs to get more familiar Assist, covering everything from basics to cross-mesh and emulation; I adapted them for public consumption, enjoy: https://t.co/EAyAsmOuhH
RT @jsrailton: 2/ Earlier this year two Israeli companies were exposed marketing ways to spy on @Starlink users, including:
☑️Locate Starlink users exact locations
☑️Identify the users, in some cases.
Reporting by @omerbenj even had screenshots from a demo given in Vienna...
https://t.co/JBBMTKaGWr
RT @calif_io: Calif researchers are at Unprompted, presenting zero-click attacks and php2shell. You don't want to miss this!
Tue Oct 27, Stage 1, 16:45, "0-Click Remote Exploits in an AI World" with @pwnlambda and @brucedang
Wed Oct 28, Stage 1, 14:45, "PHPocalypse: Finding, Exploiting, and Disclosing php2shell" with @lcfr_hax and @XorNinja
RT @NeowinFeed: Microsoft quietly removes TPM function from Pluton on 2026 AMD, Intel, Qualcomm chips.
^P https://t.co/6J4bZSeTOl
This XSS filter removes every script tag. Well, every other one. A live NodeList, a loop that removes items from it, and CVE-2020-15885.
Free lesson: https://t.co/dkEcUeK5rQ https://t.co/ptILME0SnQ
RT @vxunderground: Hello, Little People Living Inside Of My Computer,
I have made my first YouTube video. I am bonking goop (malware) with a stick. I've tried to be both technical and non-technical to appeal to all nerds.
This was a lot of work. I love you.
https://t.co/QF4yt9AuzQ
My favourite LLM harness right now is unquestionably @bot.
take this for example, I wanted it to go find every State of [Something] Security from 2025/2026 and it spent an hour just grinding through it.
At 400+ reports - this is going to be a fun week just reading through all of these to see what and how other people think.
RT @jsrailton: UPDATE: Israeli company we caught delivering zero-click infections is #1 bidder on Starlink ads in 105 countries.
Data from @trendai_RSRCH suggests worldwide surveillance op targeting @Starlink users. 1/
By @razhael
https://t.co/uWFzemhgnL https://t.co/FxkDlK7Xo0
We love grep, but there's more to security code review than searching for dangerous functions and patterns.
Those searches can give you useful starting points, but the real work is understanding what the application does, how its security controls work, and where its assumptions break down.
A missing authorisation check won't necessarily contain a suspicious keyword, and a business logic flaw might use perfectly safe functions in the wrong order.
Use grep to find places to look, but always take the time to understand what you're reviewing..
从凭据泄露到取得全局管理员时间极短,托管服务商成为风险扩大源头
Huntress researchers Jonathan Semon & Rich Mozeleski show how stolen ConnectWise credentials were used against nearly 1000 M365 tenants. In two cases, attackers gained privileged roles within three hours. https://t.co/hJmfeYb5NG https://t.co/ROJJ4nzpGv
满分漏洞已有可用利用链,未打补丁的实例大概率会被快速扫到
RT @sansecio: 🚨 BlockStrap (Magento / Adobe Commerce, CVSS 10.0)
We've reproduced several RCE chains and expect threat actors to follow suit imminently.
If you haven't patched yet, we strongly urge you to do so now.
https://t.co/7MWkyRDb4F
扫描器自动化程度高,暴露在公网的站点几乎没有缓冲窗口
RT @Unit42_Intel: We observed in-the-wild exploitation of CVE-2026-102785 on a site running a vulnerable version of YOOtheme Pro two days after the patch. Scanners planted JPEG/PHP polyglot webshells in uploads, then operators deployed an ALFA webshell. Details: https://t.co/hcKC5Te6pM https://t.co/74aAW5G8qI
继续利用硬件钱包品牌的信任投毒搜索广告,是可直接复用的手法
Vojtech Moravec at GenDigital details how AMOS can replace Ledger and Trezor apps to steal phrases. The campaign used Google Ads and fake https://t.co/S5xjJIlniu pages. https://t.co/82lUYGkQ8x https://t.co/DHc5GoHpSt
重装系统无法清除的持久化路径,对事件响应流程是明确挑战
The Trellix team found Sirius Backdoor behind a blank FedEx PDF. A provisioning package allows it to reinstall after a Windows reset. https://t.co/on54tcG7T5 https://t.co/YtnMqU7qFQ
EDR 移出内核后的新接口尚未公开文档,早期理解有先发价值
RT @elasticseclabs: Microsoft wants EDR out of the Windows kernel. We checked the binaries
They even come with public symbols. Full write-up from @zodiacon : https://t.co/9w3mFPIufM
Microsoft's answer is the Windows Endpoint Security Platform (WESP). It's undocumented, but it already ships in Windows Insider builds.
What we found:
• wesp.sys: a Microsoft-built file system mini-filter at altitude 329500
• EspFilterPort: the filter communication port it uses to reach user mode
• espclient.dll: 120 Esp* exports for rules, filters and event queues
展示 AI 在攻击流程中的实际落地环节,可对照自身防御假设
Unknown Threat Actor Uses AI-Driven ARTEX to Target South Korean Finance - @CrowdStrike
https://t.co/ftOtCFMFuD
ai_agentapt金融报告
@pentest_swissky
source ↗
把硬件层抓包与游戏客户端 RCE 串成完整链条,手法可迁移
Think reverse engineering is just Ghidra?
Gbps covers PCIe MMIO/DMA, captures PCILeech TLPs with a protocol analyzer, and breaks down a full-chain Source Engine client RCE using Frida.
This blog deserves a bookmark:
https://t.co/xKguwkVYeR
Author: https://t.co/S55mSZYIcP
#ReverseEngineering #ExploitDev #PCIe
域名影子化滥用可信站点声誉,是可观测的流量劫持手法
Infoblox traced Hidden Hawk's gambling SEO across school and government domains. Researchers contacted the group using a fake casino persona. https://t.co/MJty1sZjt2 https://t.co/sdKb2l5gQW
ok patch brotli, we just popped a shell in a client engagement, not sure if open models can write the exploit tho.
Modern SoCs have so many power optimizations. Working in performance engineering exposes us to so many interesting topics :)
P-states, DVFS, AMU, PMU, feedback loops... there’s always more to learn.
RT @Trellix: Seemingly routine threats can escalate fast, posing a larger threat to your organization.
Our Threat Response Team experienced this first hand when they were investigating an alert involving the unexpected execution of a PDF document.
Learn more: https://t.co/skPs0PaU96 https://t.co/ZgZbHBLCgD
为漏洞优先级排序降温:凭据滥用与供应链才是主要入侵路径。
I actually expect AI to make software considerably more secure in the coming years. Obvious coding mistakes will become harder to find, and exploiting software vulnerabilities may become much more difficult than it is today.
But you don’t need to discover a new software vulnerability to compromise most organizations.
Stolen or reused credentials, phishing, session hijacking, social engineering that tricks people into executing malicious code, compromised software suppliers, trojanized applications, malicious updates, insecure configurations, exposed secrets and backups, excessive privileges, abused remote management tools, and weaknesses in authentication and business processes.
There are plenty of ways into an organization that don’t require developing an exploit or discovering a new vulnerability.
And AI agents can automate and combine these methods just as effectively. They can test different approaches, adapt when something fails, and work their way towards the objective.
That’s what this post was about.
提示防御投入应偏身份与社工面,而非只盯 CVE 数量。
People seem to think this post was about exploit development. It wasn’t.
You don’t need to discover a new software vulnerability to compromise 90% of organizations.
展示 agent 自主执行长链任务的能力边界,关系到其可授予的权限。
I was waiting for some research to complete and I decided to let astra have a look at some reinforcement learning work I did as a side project early 2025. Astra was very generous when it evaluated my novice attempts at training.
I told it to do whatever it liked and retrain from scratch following the original constraints set out. After 30 minutes it came back and, well, you can see the results for yourself.
可作为自托管数据管控组件的候选,成熟度需自行评估。
oak — Meaningful control of data in distributed systems. https://t.co/FINrLk9CBR https://t.co/TMZa7RqEIw
涉及 LLM 推理成本结构,影响自建与调用模型的费用估算。
Is the idea dynamic output token pricing? Who is going to agree to that?
反映本地评测模型能力时的算力与费用门槛。
i want to run lot of these benchmarks to understand model capabilities more broadly, but man its so expensive.
自建站点被 AI 训练爬虫压垮资源的实例,可参考限流与封禁策略。
以前是claude爬虫疯狂爬我的网页,现在meta来疯狂爬取了,频率很高,影响到服务器正常运行了😅
可能是样本或镜像校验值,需回溯原始上下文才能实际使用。
8453fffa759e6ad44f19d58d26afa56380c3d3d1daef49f1771dff38f0b8167e ca80443f943911bb75a6dc04e0027234dfbc4c1fcc0d7ef5574f8818bd7bb6b5
自建 Linux 环境的攻击面连载,附可复现代码与思路
https://t.co/TXOcChZaVg next one from my blog #linux #hacking series! enjoy ❤️
https://t.co/Uyfk8jpbtv
#research #malware #programming #rootkits #rootkit #purpleteam #cybersec #cybersecurity #book https://t.co/qFTMMRsRpl
直接关系到选本地还是云端模型来跑攻击性 Agent 的取舍
Check out our latest benchmark comparing open and closed models on real-world offensive cyber capabilities.
👇
异常账号与登录类型 3 的完整排查链,可直接对照横向移动检测
The first thing that looked strange was not on the user’s laptop.
It was on the file server.
A workstation called WS-214 suddenly authenticated to our Finance share using an account called:
svc-reporting
That account was supposed to belong to an old reporting service.
It had no reason to be coming from a normal employee workstation.
So I checked the Security logs on the file server.
Event ID 4624 showed a successful network logon.
Logon Type: 3
The account was CORP\svc-reporting, and the source was WS-214.
A few seconds later, Event ID 5140 showed the same account accessing:
\\FS-08\Finance
Now I wanted to know what happened on the workstation before that connection.
The process logs showed PowerShell launching cmd.exe, and the command line included a net use connection to the Finance share using the service account.
That was enough to treat it as possible lateral movement.
I checked the account in Active Directory.
Its description still said:
Legacy reporting service
The password had not been changed since February 2025.
That told us something else.
This was an old service credential that had been sitting around for a long time.
We isolated WS-214, disabled the account, rotated the credentials and started checking every system that had used svc-reporting.
We also reviewed the Finance share to see what had been accessed during the session.
What stood out to me was that the attacker did not need some complicated exploit to move between systems.
They had a valid credential.
Windows accepted it.
SMB accepted it.
From the network’s point of view, the login looked legitimate.
That is why stale service accounts can be dangerous.
They normally run quietly for years, people forget who owns them, password rotation gets skipped, and eventually nobody is really watching where they are being used from.
Sometimes lateral movement does not look like an exploit.
It looks like a successful login from the wrong machine.
dfir横向移动windows
@Officialwhyte22
source ↗
现成的载荷隐藏与加载手法,可用于免杀与检测规则验证
RT @ipurple: PngLoader - hides arbitrary data inside PNG images using two layers:
1⃣ Nibble-to-ASCII encoding - binary → printable ASCII characters
2⃣ Fibonacci steganography - encoded data → scattered pixels in a PNG
3⃣ Loader - a Windows "PNG viewer" that extracts and runs the hidden data
https://t.co/KOK78v1npo
自建远控管理面本身就是高价值入口,需评估暴露面与鉴权
MeshCentral — A complete web-based remote monitoring and management web site. Once setup you can install agents and perform remote desktop session to devices on the local network or over the Internet. https://t.co/f69NvmVDH7 https://t.co/TBaRlv61Dl
presume compromise, not assume breach
RT @gdb: Announcing three major scientific discoveries, one each in mathematics (harmonic analysis), computer science (complexity theory), and physics (quantum field theory).
AI starting to help us improve our understanding of the universe! https://t.co/VkfacopR2R
RT @loktar00: @sfxnz That's the cool thing though, you can just restrict access locally. Whereas cloud you have 0 choice or power.
RT @sfxnz: So running local models isn’t really private anymore…
RT @CyphrRiot: @0xTMK Don't buy blind.
Identify the exact chip first (`lspci -nn | grep -i network`). OpenBSD's `bwfm(4)` supports some PCI Broadcom FullMAC chips (BCM4350, 4356, 43602, 4371, 4378, 4387) after `fw_update bwfm`
https://t.co/esK1K4OG8L https://t.co/qOi9XasSZ0
RT @CyphrRiot: OpenRiot v8.0.7 released. Minor speed improvements. Better network handling. New Backgrounds.
Start a Riot. https://t.co/tAXPwGj6UC https://t.co/SByzYohQC2
移动端内核 UAF 到 root 的完整链路拆解,可复用于同类漏洞利用。
RT @eshard: Kernel UaF to Root, part 3:
Turning the address we leaked in part 2 into a kernel write, on a Pixel 6 running Android 12.
🔗Read: https://t.co/nzZ4vFCHmC
新开源 C2 框架落地后会快速进入实战与检测对抗,需提前掌握其能力边界。
reminder, blackmagick c2 release halloween night. full open source for https://t.co/zKV2rbHvmH pro users. i’ll explain how it all works, show how to extend it, and provide the tools to do so. livestream halloween night. hand out candy then get comfy and join the stream.
前沿模型独立完成 PHP 到 shell 的利用链,反映自动化攻击能力的水位。
RT @qwerty__cs: php2shell挑战赛
胜出者只有Claude Mythos 5.1和GPT-5.6 Sol https://t.co/dv8LeLl5hU
It's very certain AI cloud models go over all the cache and they know when you hit, they cache it.
On the cache they work faster, they collectively share all the bugs to everyone, the question is do they tell you.
It's a collective white hat effort
把 LLM 当恶意代码阅读器是低成本、可复现的样本分析路子
When I was getting started with malware - I praised Zeus and Carbanak (namely Carbanak) as teaching me a lot of what I knew.
I remember spending days just reading through the source and reimplementing stuff to get in the head of the developer.
I'm surprised I never decided to take a different look at it with an LLM.
Well I know what I'll be doing this morning...
解锁后可得超出官方限制的显存与算力,直接影响本地跑大模型的门槛
RT @KBlueleaf: Now thanks to community resources like gpu-unlock-ada or gpu-unlock-4080s. I successfully unlocked my 4090 on "Windows" system.
All the details are released in https://t.co/UOfas5elK9
Plz check it out!
https://t.co/i4exHAkipR https://t.co/pbPAmvTGFW
RT @calysteon: CVE-2026-65373, CVE-2026-112511, CVE-2026-84496, CVE-2026-113946, CVE-2026-113941, CVE-2026-84498, CVE-2026-84590, CVE-2026-112482, CVE-2026-112457, CVE-2026-112443, CVE-2026-112708, CVE-2026-112461, CVE-2026-112417, CVE-2026-112481, CVE-2026-112451, CVE-2026-112622, CVE-2026-112529, CVE-2026-65384, CVE-2026-112526, CVE-2026-112620, CVE-2026-112671, CVE-2026-112519, CVE-2026-112414, CVE-2026-112604, CVE-2026-112546, CVE-2026-112413
RT @offensive_con: Keynote - Recursively self-improving security: AI offense, AI defense, and what's left for the humans by @oegerikus https://t.co/SpX9Yde3aS