Megatron LLM Hub
All digests
Twitter

推特安全流

Top6

SpecterOps 开源 79 个红队技能、22 个 Agent、26 个插件族,供 AI Agent 调用。

首个将成熟红队能力系统化封装为 AI Agent 技能的开源库,直接推动 Agent 化攻击研究。

Infosec tradecraft just became reusable by AI agents. SpecterOps just open-sourced: 79 skills. 22 reusable agents. 26 plugin families. BloodHound. Cobalt Strike. Outflank C2. Ghidra. Binary Ninja. Ghostwriter. Recon. AppSec. Code review. C2 development. Reverse engineering. Adversary simulation. Windows + macOS tradecraft. And this is NOT just for red teamers. Vulnerability researchers: These workflows could accelerate code review, patch analysis, 1-day research and potentially help with 0-day discovery when paired with real research expertise. Reverse engineers + malware analysts: Give agents structured workflows, references and tooling instead of starting every investigation from a blank prompt. Blue teams + detection engineers: Study the same offensive tradecraft, emulate attacker behavior, build better detections and start asking what telemetry survives increasingly agent-assisted operations. DFIR + threat intel: Understand what adversaries may automate next and turn repeatable investigative knowledge into reusable workflows. Red teamers: BloodHound attack paths, recon, C2 development, adversary simulation and operator tradecraft are becoming increasingly agent-assisted. This isn't another collection of AI prompts. It's practitioner knowledge being turned into reusable, reviewable security workflows. Potentially useful for everyone from CTF learners and newcomers all the way to malware analysts, reverse engineers, exploit devs, red teams, blue teams and vulnerability researchers. This is only the beginning. @SpecterOps Skills: https://t.co/Cj77Ir3Qlj @OutflankNL and @kyleavery breakdown: https://t.co/q5sEaZlp1E #Infosec #RedTeam #ReverseEngineering

ai_agent工具红队 @cr3ghost source ↗

CrowdStrike Falcon 0day LPE 漏洞 FalconFlank 已公开,含 PoC。

主流 EDR 产品 0day LPE,PoC 公开,对红队与防御方均有直接参考价值。

RT @MSNightmare2000: FalconFlank : Crowdstrike Falcon 0day LPE is now public https://t.co/J0oBnyVkH5

lpeedrpoccve @pentest_swissky source ↗

CrowdStrike Falcon 0day LPE 漏洞 FalconFlank 已公开,含 PoC。

主流 EDR 产品 0day LPE,PoC 公开,对红队与防御方均有直接参考价值。

Another fresh CrowdStrike Falcon 0-day dropped from @MSNightmare2000. This time the target appears to be CrowdStrike Falcon. FalconFlank. Red teamers: Want to bypass EDR? Pay attention to what this means for operating around modern EDR. Detection engineers / blue teams: public research against endpoint security products has a habit of turning into real tradecraft very quickly. Malware analysts and threat intel: this is exactly the kind of PoC worth understanding before it starts showing up elsewhere. https://t.co/DPYQNMocy1 #EDR #RedTeam #DetectionEngineering #Infosec

lpeedrpoccve @cr3ghost source ↗

Astra 自主漏洞研究 Agent 在多个加固目标上发现并利用多个 0day。

AI Agent 自主完成漏洞发现与利用,是首要关注意图的核心进展。

RT @ky1ebot: We evaluated Astra's cyber capability with minimal orchestration. Astra was able to perform long-horizon vuln research, find and exploit multiple 0day vulnerabilities in multiple hardened targets autonomously with minimal to no human interaction. https://t.co/QMC7PCTbb7

ai_agent0day漏洞研究 @moyix source ↗

XBOW 团队完成 Chrome 全链漏洞利用,获 Google 25 万美元奖励。

AI 驱动的漏洞研究团队达成 Chrome 全链利用,标志性成果,影响浏览器安全研究。

RT @oegerikus: Incredible, and a little frightening! In January at the @XBOW all-hands in Malta, I set @moyix and his team the challenge of a full chain exploit. I never dreamt they'd get there so quickly. To my knowledge, it's only the 2nd time ever Google has paid out $250K for one of these.

ai_agent浏览器漏洞利用 @moyix source ↗

PyPI 出现针对 requests 库的 typosquatting 恶意包,窃取密钥并反弹 shell。

供应链攻击,影响广泛使用的 Python 库,需立即排查依赖。

RT @anyrun_app: 🚨 Typosquatting packages on PyPI are targeting 𝗿𝗲𝗾𝘂𝗲𝘀𝘁𝘀, one of the most widely used Python packages. Catch them before compromise. Attack timeline: 3 hours ago, a new PyPI account was registered. One hour later, four packages were published: 0requests, py-0requests, py-1requests, and py-2equests. ⚠️ On import, each package: – collects env vars prefixed SECRET / API / TOKEN / KEY – sends them over a raw TCP socket – spawns a reverse shell hook The callback currently points to 127.0.0.1, so the payload is staged, not armed, but the host is read from TS_HOST. One update could switch exfiltration to a live C2 channel and turn it into an active supply-chain threat ❗️ 👨‍💻 Full behavior analysis and IOCs in #ANYRUN Sandbox: https://t.co/wmT4bjqbo5 📌 All packages on PyPI: https[:]//pypi[.]org/user/preet780/ A mistyped dependency in a CI/CD pipeline can be enough to leak credentials. Catch it before the package becomes a real compromise with #ANYRUN: https://t.co/rc55Cne7cF

供应链恶意包pyPI @0x534c source ↗
Must-see3

XBOW 的 Native 团队获得今年首个 Chrome 全链漏洞利用奖金。

AI 驱动的漏洞研究团队达成 Chrome 全链利用,标志性成果,影响浏览器安全研究。

RT @Xbow: This just in: XBOW's Native Team has claimed the first Chrome Full Chain Exploit Bonus this year. Only four will be awarded in 2026. ⚡️ https://t.co/JYpPnZtesi

ai_agent浏览器漏洞利用 @alkalinesec source ↗

Agentic 逆向工程去混淆技术公开演讲录像与幻灯片已发布。

展示 AI Agent 在逆向工程中的实际应用,对恶意软件分析有参考价值。

RT @mr_phrazer: The recording of "Deobfuscation in the Age of Agentic Reverse Engineering" is now public: https://t.co/jMO0GHRM8l We (CC @nicolodev) show how to use agents to break protections found in anti-cheats, DRM systems & commercial protectors. Slides: https://t.co/Fz2DhU2RlT

ai_agent逆向工具 @_winterknife_ source ↗

开源 MCP 实现端到端 agentic 检测工程,从报告到查询自动化。

将 AI Agent 应用于检测工程,是蓝队自动化的重要实践。

RT @M_haggis: Anyway - saw something about automating DE work and whatnot, reminded me of my MCP- https://t.co/UUmmsT1ERR Full end-to-end agentic detection engineering is very possible. From report -> new/current analytics -> simulate with Atomic Tests (new/current) -> query SIEM and iterate on query -> open PR. Shipment complete. Full vid here - https://t.co/x27o36J9k5 Cursor Subagents, LangGraph - all there. Hope it helps someone out there. The MCP is an awesome way to get past the whole "AI can't write good queries" or "AI can't do DE work" - question everything but also POC it and prove it out.

ai_agent检测工程工具 @ipurple source ↗
Recommended15

Windows 预览版将支持 agentic app containers,可任意完整性级别运行。

新容器机制可能引入新的攻击面,与 AI Agent 安全直接相关。

RT @yarden_shafir: Preview builds are adding support for "agentic app containers": AppContainers that can run with any integrity level, have access to more resources than classic AppContainers and be virtualized through bfs.sys as if they were application silos https://t.co/xl39ONAAS0

ai_agentwindows容器安全机制 @kernullist source ↗

Hugging Face 发生安全事件,社区讨论激烈。

AI 基础设施平台的安全事件,对依赖其服务的用户有直接影响。

Going to start unnerving people by going around calling it the “First Huggingface Incident” and refusing to explain

ai_agent供应链事件 @moyix source ↗

OpenAI Hugging Face 事件独立报告作者发布 AI 安全风险分析。

提供对 Hugging Face 事件的深度分析,有助于理解 AI 供应链风险。

RT @ZackKorman: Totally normal post about AI safety / risk by one of the authors of the independent report on the OpenAI Hugging Face incident. https://t.co/vj0mUsOiRO

ai_agent供应链报告 @Teach2Breach source ↗

Quarkslab 发布 LLVM 混淆与去混淆分析文章。

深入编译器优化与混淆对抗,对逆向与恶意代码分析有价值。

LLVM based analysis of the relationship between compilers, obfuscation, and de-obfuscation https://t.co/7rfwhEczcz Credits Robert Yates (@quarkslab) #infosec https://t.co/U2hcoS0igE

混淆逆向议题 @0xor0ne source ↗

发布 KQL 检测规则,用于识别 Teams IT 帮助台冒充钓鱼。

针对 Teams 钓鱼攻击的检测规则,对蓝队有直接实用价值。

🔍 Detect Teams IT Helpdesk Impersonation Msg Phishing & Vishing Human-operated intrusion campaigns abusing Microsoft Teams external collaboration to socially engineer users into granting interactive remote sessions are on the rise. To help defend against this vector, I've developed a unified KQL detection to surface Initial Access via Teams Chat and Voice (MITRE ATT&CK T1566.003). https://t.co/uPL4t1bjdj #Threathunting

检测工具钓鱼 @0x534c source ↗

Malleon 工具可自动化生成 Cobalt Strike Malleable C2 配置。

自动化 C2 配置工具,对红队有实用价值。

Malleon - automates HTTP/HTTPS configuration for Cobalt Strike Malleable C2 profiles using real traffic captured from legitimate applications https://t.co/ONhteZwbMz

工具c2红队 @ipurple source ↗

朝鲜黑客在 macOS 恶意软件中植入虚假错误信息以逃避 AI 分析。

针对 AI 安全分析工具的对抗技术,对 AI 安全研究有参考价值。

RT @TakSec: Attackers plant fake errors inside malware to evade AI analysis North Korea-linked actors used a macOS implant @LabsSentinel calls “Gaslight”. How it worked: 1. Create a macOS backdoor in Rust 2. Add 38 fake “system” messages 3. AI security tooling reads the file 4. The AI model percieves system failures 5. The AI model stops analyzing 6. The actual malware continues executing The malware also contains a credential stealer, interactive shell, and Telegram C2. Unlike the nuclear weapon technique, this did not target safety guardrails. It targeted the model’s perception of its own system state. Full write-up in the comments. 👇

ai_agentmacos恶意软件 @0xocdsec source ↗

CouchPotato 提权工具发布,利用 SeImpersonate 提权至 SYSTEM。

Windows 提权工具,对红队有实用价值。

RT @_MrNiko: 🚨 another potato just dropped CouchPotato: SeImpersonate → SYSTEM ETW + AMSI patched. indirect syscalls. service account or admin → NT AUTHORITY\SYSTEM https://t.co/1ODkC0yIn9 #ADSecurity #ExploitDev #InfoSec

提权工具windows @0xocdsec source ↗

过度防护的 AI 模型正将蓝队推向开源模型。

讨论 AI 模型安全限制对蓝队工作流的影响,与 AI Agent 安全相关。

RT @attrc: Overzealous guardrails are going to permanently push the blue-team cybersecurity community toward open-source models. If defenders can’t reliably use frontier models for legitimate security work, they’ll build their workflows somewhere else and not look back. This is already occurring.

ai_agent蓝队议题 @unpacker source ↗

Praetorian 发布 BLE 安全测试工具 Caeruleus。

BLE 安全测试工具,对 IoT 安全研究有实用价值。

Bluetooth Low Energy Security Testing, Consolidated: Introducing Caeruleus - Aaron Wasserman, Hunter Ver Helst, Siddhant Kalgutkar, Will McCardell - @praetorianlabs https://t.co/ucfCFQSaoJ

工具iotble @pentest_swissky source ↗

Harden-Windows-Security 工具更新,使用官方方法加固 Windows。

Windows 安全加固工具,对系统管理员有实用价值。

Harden-Windows-Security HardenSystemSecurity-v1.0.80.0 — Harden Windows Safely, Securely using Official Supported Microsoft methods and... https://t.co/lj4OgE8aW6 https://t.co/JGeXp7VPri

工具windows加固 @KitPloit source ↗

smolvm 便携式轻量虚拟机工具更新。

轻量虚拟机工具,对恶意软件分析有实用价值。

smolvm v1.8.3 — Portable, lightweight, self-contained virtual machine. https://t.co/WJzQENvZJS https://t.co/0uxlzfXLaP

工具虚拟机恶意软件分析 @KitPloit source ↗

warpgate 堡垒机工具更新,支持 SSH、K8s、数据库等协议。

多功能堡垒机工具,对基础设施安全有实用价值。

warpgate v0.28.0-beta.1 — Fully transparent SSH, HTTPS, Kubernetes, MySQL and Postgres bastion/PAM that doesn't need additional client-side software https://t.co/IziPauQgSn https://t.co/5XQkvmv5J8

工具基础设施加固 @KitPloit source ↗

Cowrie SSH/Telnet 蜜罐工具更新至 v3.0.12。

蜜罐工具更新,对威胁情报收集有实用价值。

cowrie v3.0.12 — Cowrie SSH/Telnet Honeypot https://t.co/kLW08X4QGh https://t.co/3G5YnlmnDd https://t.co/WE3BrmAfeM

工具蜜罐威胁情报 @KitPloit source ↗

watermarks-remover 工具可去除多种 AI 生成内容水印。

AI 内容水印去除工具,对 AI 安全研究有参考价值。

watermarks-remover v0.5.0 — Strip multi-vendor AI provenance marks: Unicode text hygiene, statistical rewrite hooks, and C2PA/metadata... https://t.co/dLB3fdE72I https://t.co/vDc7upPyrz

ai_agent工具水印 @KitPloit source ↗
More166

Nvidia 以 130 亿美元收购 Hugging Face。

AI 基础设施重大并购,可能影响模型托管与分发生态。

RT @arstechnica: Nvidia buys Hugging Face, the Github of AI, for $13 billion https://t.co/weJOjYtxNG

ai_agent供应链事件 @artem_i_baranov source ↗

四大主流 AI 模型遭遇罕见的同时段宕机。

AI 服务可用性事件,可能影响依赖这些模型的应用。

RT @arstechnica: Four major AI models suffer rare overlapping downtime https://t.co/PWdIPJjG9h

ai_agent事件可用性 @artem_i_baranov source ↗

新 loader 利用 .reloc 节和反虚拟机检查分发 CASTLESTEALER。

恶意软件加载器技术分析,对检测与防御有参考价值。

Lost in relocation: analysis of a new loader distributing CASTLESTEALER analyzed by @DanielStepanic @k33b0i. TLDR:- New loader evades static detection using .reloc section abuse, five anti VM checks and MBA obfuscation to deliver via Google Ads. https://t.co/KKYtaZERIX https://t.co/pNlDIsj9WE

恶意软件loader报告 @5mukx source ↗

被劫持的 YouTube 频道推广新型信息窃取器 REVSTEALER。

新型窃密恶意软件分析,包含沙箱规避与 App-Bound 绕过技术。

Hijacked YouTube channels are pushing an emerging infostealer. We're tracking #REVSTEALER: at least 17 compromised channels advertising free game cheats, funneling viewers to 2 malicious domains. The stealer itself is built for long-term operation: - Sandbox scoring: 10 weighted checks, self-terminates at a score of 7+ - App-Bound Encryption bypass: reads the decrypted key from browser memory under debugger control - EtherHiding: fallback C2 stored in a Polygon smart contract, swappable without touching the binary - Gaming focus: session cookies from Steam, Battlenet, and Roblox, feeding account resale markets - 4,700 related samples hit VirusTotal in the past year. Our team observed 4 follow-on modules that haven't been publicly documented until now. Full analysis, whitepaper, YARA rule, and an IDAPython string decryption tool by @k33b0i, @soolidsnakee and @DanielStepanic: Blog: https://t.co/4jGSZ2Tidg Whitepaper: https://t.co/P8yqn0uCdT

恶意软件窃密报告 @elasticseclabs source ↗

ClickFix 攻击活动开始针对 macOS 用户。

macOS 平台 ClickFix 攻击,对 Mac 用户安全有参考价值。

RT @ValeryMarchive: This ClickFix campaign targets macOS users. - press29[.]com is used to host the (2) shell scripts as well as the final payload. - render-6[.]com appears to be used for statistical purposes in the second script. The final payload is here: https://t.co/FhutdJsQxl

macos恶意软件事件 @patrickwardle source ↗

攻击者滥用 Teams 外部协作,部署 Node.js 植入物进行持久化。

针对 Teams 的社会工程攻击,对防御方有检测参考价值。

RT @MsftSecIntel: Microsoft Threat Intelligence is tracking a human-operated intrusion campaign in which attackers are impersonating IT personnel & abusing external Teams collaboration to gain remote access and deploy a Node.js implant for persistent command execution & C2. https://t.co/DXLX2CVukP After establishing access, the attackers use trusted tooling to perform reconnaissance, capture screenshots, execute follow-on payloads, and move laterally toward domain controllers, certificate authorities, and other high-value systems. Organizations should restrict Teams external access to trusted domains, reinforce user education, and harden systems against social engineering. Read the blog for analysis, Microsoft Defender coverage, indicators, hunting queries, and mitigation guidance.

事件社会工程恶意软件 @virusbtn source ↗

Huntress 发现恶意 ScreenConnect 客户端传播 VBScript 蠕虫。

RMM 工具被滥用,蠕虫式传播,对使用 ScreenConnect 的组织有警示。

RT @HuntressLabs: NEW RMM TRADECRAFT: In the last few weeks Huntress has observed a notable pattern across multiple customer environments: rogue ScreenConnect clients repeatedly spawning the Windows Script Host to execute a series of four Visual Basic Script files. https://t.co/GPT01j8MVH

事件恶意软件rmm @virusbtn source ↗

Huntress 观察到恶意 ScreenConnect 客户端自动传播 VBScript 文件。

RMM 工具被滥用,蠕虫式传播,对使用 ScreenConnect 的组织有警示。

Huntress researchers observed a worm-like activity with rogue ScreenConnect RMM clients: previously installed, modified ScreenConnect clients can automatically transfer and execute the same four VBScript files on newly connected ScreenConnect endpoints. https://t.co/Luh9Hde3qA https://t.co/VH7MB6AwIq

事件恶意软件rmm @virusbtn source ↗

BoratRAT 利用类似 ClickFix 手法,通过 DNS TXT 获取载荷。

恶意软件传播技术分析,对检测与防御有参考价值。

RT @abuse_ch: #BoratRAT spreading using similar tactics as #ClickFix 👇 1️⃣ Fake Microsoft Security Verification 🔑 leading to malicious PowerShell execution 🖱️ 2️⃣ Command triggers a DNS TXT request to recapture-robot .today 🌐 to obtain a PowerShell script 📜 3️⃣ Script drops payload, infecting host with BoratRAT 💻 Payload delivery: 🌐 recapture-robot .today (PDR 🇮🇳) 🌐 91.193.7.186:49094 (M247 🇯🇵) Botnet C2: 📡 out-agent.duckdns .org 📡 91.193.7.186:48988 (M247 🇯🇵) 📄 Sample: https://t.co/7xbU0TCvNv

恶意软件rat事件 @virusbtn source ↗

PhantomStealer 新变种增加文件加密功能,密钥上传至 FTP。

窃密软件向勒索软件演变,技术细节值得关注。

RT @Threatlabz: ThreatLabz has identified a new PhantomStealer variant that contains code to perform file encryption in addition to credential theft. The ransomware feature is configured at build time and uses 256-bit AES in CBC mode. Unlike most ransomware families, the AES parameters are not locally encrypted with an asymmetric algorithm. Instead, this PhantomStealer variant uploads the encryption parameters including the AES key, IV, and an HMAC to a preconfigured FTP server. This campaign is targeting corporate users through business-themed spear phishing emails. The infection chain begins with a malicious ZIP attachment containing an obfuscated JSE file that downloads a second-stage PowerShell script. The script loads PhantomStealer in memory (without touching the disk) to evade antivirus and endpoint detections. The following IOCs are associated with this campaign: Payload URL: cns-fl[.]com/dv/secured_stub.ps1 C2 mail server: mail.trimnt[.]com Email accounts: phan@trimnt[.]com and phan2@trimnt[.]com SHA256 file hashes: 3b366ac5bae87508a67dc2a4c6c452aecc6e3af54210eef77f9d9d1a3d18a9ed (JSE) C01086d1963c687ece7cb41e9e96e6700b0b4b60483187a37011144f4ac5f756 (JSE) 94285e38119a565200acfed9454d68f70f31a58701633ac2f45c844024a568ad (zip) Cbb573d028ee4b2550568e76b5e651ab33168bc7a4ecb41a26585d6804aeabd2 (zip) 97d58b8fc012ccf58865d0fff86a732034618e490d9be5f3bc7e54c154661f20 (zip) 854075c837e5e9698ded7f09ca416a40b12410a7b78bfaa29adef92a18a97545 (tar) 1eca9708f836b9126c408fdd1b3c88b6f7292c23953a87d5259e37cd2c647ab5 (ps1) 21d5da881572ecf5a28d27640091617a591f89f395a1a3b6154de440964832bb (dll) 0336799b651f55b0ff0c9561509db6e66a5a6b160c9e38c885f8e3367ce94522 (exe)

恶意软件勒索报告 @virusbtn source ↗

Check Point 揭露针对巴西组织的 SEO 操纵活动。

SEO 操纵攻击案例分析,对防御方有参考价值。

Check Point's Amit Yardeni uncovers a sustained SEO manipulation campaign against Brazilian organizations by Gambling Goblin: a Chinese-speaking cybercrime cluster connected to the Earth Berberoka group, that targeted gambling sites across Asia. https://t.co/uVYT52pYoy https://t.co/vQFJlXgL9w

事件seo报告 @virusbtn source ↗

StreamRat 安卓银行木马通过 Meta 和 TikTok 广告传播。

移动端银行木马传播方式分析,对移动安全有参考价值。

ThreatFabric researchers analyse StreamRat, a new Android banking trojan promoted to Spanish-speaking users through Meta and TikTok advertisements impersonating a free TV-streaming service. https://t.co/6o6wvOfQ6G https://t.co/rTMVbpKWhV

移动端恶意软件报告 @virusbtn source ↗

8kSec 发布 Android SELinux 内部机制系列文章。

深入 Android SELinux 内部,对移动安全研究有参考价值。

Series on Android SELinux internals (@8kSec) Part 1: https://t.co/r8RhF2WIhE Part 2: https://t.co/57bpqR5IgM Part 3: https://t.co/SOOqLt0Xdq Part 4: https://t.co/7DcJeMuol5 #infosec https://t.co/C5JYBnULg7

移动端selinux议题 @0xor0ne source ↗

Fedora openvswitch UAF 漏洞 CVE-2026-31678 利用代码公开。

Linux 网络组件漏洞利用,对相关系统管理员有参考价值。

RT @cybermeowfia: Today's exploit is for the latest Fedora, a UAF in openvswitch: CVE-2026-31678. It was introduced in Aug 2015 and fixed upstream in Mar 2026. Discovered and exploited by the NebuSec security pipeline. Exp source code: https://t.co/feNDoZ8rcM https://t.co/LApCtejKOq

cvepoclinux @0xocdsec source ↗

NVIDIA Jetson 存在 Secure Boot 绕过漏洞,需修补并熔断密钥。

硬件安全漏洞,影响使用 Jetson 的设备,需及时处理。

RT @qkaiser: Secure Boot Bypass on NVIDIA Jetson, it’s time to patch and fuse those keys ! https://t.co/qo2rEtI0Mp https://t.co/5eJHQwt9R2

固件cve硬件 @0xocdsec source ↗

WhatsApp 漏洞仅影响 Android,iOS 不受影响。

明确漏洞影响范围,对移动端用户有参考价值。

RT @androidmalware2: Is iPhone affected? I have tested yesterday's published WhatsApp vulnerability against latest iOS and it is not affected This WhatsApp security issue affects only Android devices where it allows to access photos during a WhatsApp video call on PIN locked smartphone https://t.co/vUqQeBgoEG

移动端cve事件 @0xocdsec source ↗

Elastic 发布 Linux 无文件攻击执行模式分析。

深入 Linux 无文件攻击检测,对蓝队有参考价值。

"Fileless" doesn't mean invisible on Linux. We examine five execution patterns, their observable behavior, and the opportunities defenders have to detect them. Research by @RFGroenewoud: Blog: https://t.co/OjTOLTA4UE Helper: FENIX can be used to learn and check coverage in a lab.

linux恶意软件报告 @elasticseclabs source ↗

N0va 钓鱼工具包利用 Device Code 钓鱼攻击欧美组织。

新型钓鱼工具包分析,对防御方有参考价值。

RT @anyrun_app: 🚨 We uncovered #N0va, a new phishkit targeting organizations in North America and Europe across government, technology, consulting, and healthcare. ❗️ It splits its infrastructure across compromised legitimate sites and shared/cloud hosting, including Cloudflare Workers and Linode Object Storage. Observed lures: Microsoft Security, Teams, SharePoint, OneDrive, DocuSign, Google Drive, Dropbox, Zoom, Adobe Sign. ⚠️ N0va uses Device Code phishing to obtain access and refresh tokens through legitimate Microsoft authentication, then moves toward PRT-based SSO access via token exchange and device registration. The risk is fragmented visibility: no single layer shows the full attack. 👨‍💻 #ANYRUN Sandbox helps reveal browser behavior, extract IOCs, and pivot to related activity faster. Review the analysis sessions and improve your detection coverage: 🔹 Microsoft-themed lure: https://t.co/xaV5IuAjJS 🔹 DocuSign-themed lure: https://t.co/cZB106qvpk 💬 Find IOCs in the comments and let us know which detection layer would give you the earliest signal. 🔍 Track N0va activity and uncover related infrastructure in TI Lookup: https://t.co/aVno9dNKoi ⚡️ Strengthen your SOC, detect complex threats faster, and boost team performance with #ANYRUN: https://t.co/FbaYwM84XU

钓鱼工具事件 @virusbtn source ↗

ExfilSquad 勒索团伙针对英国公共部门发起攻击。

勒索软件团伙活动分析,对防御方有参考价值。

RT @BushidoToken: New Blog! 🇬🇧 UK Cybercrime Journal: ExfilSquad Emerges - ExfilSquad’s extortion campaign targets UK public sector orgs, education, and law enforcement - ExfilSquad's primary attack vector involves exploiting CRM platforms and Microsoft Power Pages 🔗 https://t.co/EaGA63Firi https://t.co/aOWvVP3AxK

勒索事件报告 @virusbtn source ↗

ESET 报告 QR 码钓鱼在 2026 上半年达到创纪录水平。

QR 码钓鱼趋势分析,对防御方有参考价值。

RT @ESETresearch: QR code phishing – also known as #quishing – reached record levels in ESET telemetry in H1 2026 as attackers exploit the widespread adoption of QR codes in everyday life. The technique is evolving rapidly in terms of automation, scalability, and detection evasion. 1/5 https://t.co/6dvVl8wS5i

钓鱼报告事件 @virusbtn source ↗

Gen 分析 Phantom Deal 虚假收购欺诈活动。

商业欺诈案例分析,对防御方有参考价值。

Gen's Martin Chlumecký & Luis Corrons look into Phantom Deal, a fake acquisition fraud campaign. Attackers posed as executives, moved conversations to WhatsApp and personal email, and forged acquisition documents to set up international wire transfers. https://t.co/yqdGjoeX9G https://t.co/pvhxWCPsIA

事件欺诈报告 @virusbtn source ↗

FortiVPN 0day 漏洞利用在暗网出售,要价 2 BTC。

企业 VPN 0day 在暗网流通,需关注自身资产风险。

RT @ptdbugs: Sale of a FortiVPN 0-day exploit For informational purposes only. FortiVPN is among the most widely used enterprise VPN solutions, particularly in organizations using Fortinet infrastructure. The vendor’s solutions have the strongest presence in the Middle East, Europe, and Africa. Price: 2 BTC (~78K$) #dbugs_darkweb

cve0day事件 @0xocdsec source ↗

Coruna iOS 漏洞利用工具包 C2 面板分析。

iOS 漏洞利用链分析,对移动安全研究有参考价值。

RT @1ZRR4H: 🚩 Coruna iOS exploit kit An overview of the C2 web panel 👁️ iPhone → Web/Safari → WebKit RCE → PAC bypass → Sandbox Escape / Kernel Privilege Escalation → Payload. Interesting campaign highlighted in the tweet below. https://t.co/HaXIQkrHjV

移动端漏洞利用报告 @0xocdsec source ↗

CVE-2025-31200 和 CVE-2025-31201 疑似为 NSO Group 的零点击漏洞。

零点击漏洞归属分析,对移动安全研究有参考价值。

RT @billmarczak: Seems like CVE-2025-31200 and CVE-2025-31201 were NSO Group! Most recent zero-click chain that appears to be attributable to them https://t.co/LpqqV2H6Om (Ref: https://t.co/wi5X1I1iVs)

移动端0day事件 @0xocdsec source ↗

分享漏洞挖掘方法论,强调没有秘密技巧。

漏洞挖掘方法论分享,对安全研究人员有参考价值。

posted my methodology a while back, still highly relevant, no secret tricks. https://t.co/Y5eJynIuln

漏洞研究议题 @0xAsm0d3us source ↗

安全研究员因单个漏洞提交获得 10 万美元漏洞赏金。

高额漏洞赏金案例,对漏洞研究有激励参考。

I was awarded a $100,000 bug bounty for a single vulnerability submission. https://t.co/Yw4Dg38zq8

漏洞研究事件 @0xAsm0d3us source ↗

公开 LPE 漏洞挑战,要求将其转化为可利用的提权漏洞。

LPE 漏洞研究挑战,对漏洞利用研究有参考价值。

RT @portbuster1337: i have a challenge for you people, turn this LPE issue into a working LPE exploit which in turn launches a system level shell, winner gets a kiss https://t.co/bPEbia9OKU

lpe漏洞研究议题 @kernelstub source ↗

librepods 工具可让 AirPods 脱离苹果生态使用。

硬件逆向工程成果,对硬件安全研究有参考价值。

librepods nightly-53679cc — AirPods liberated from Apple's ecosystem. https://t.co/oHffRgWssw https://t.co/e4mMXpyyKS

硬件逆向工具 @KitPloit source ↗

TraceRay 函数逆向工程教程,用于游戏作弊开发。

游戏逆向工程教程,对逆向学习有参考价值。

🎯 TraceRay Function Reverse Engineering Learn how to locate and analyze the engine's native raycasting to check visibility between players. Learn to trace lines through 3D geometry to verify if a target is behind cover before executing aimbot logic. 👉 https://t.co/2nWcD4KNgT https://t.co/cPAYBavEbZ

逆向游戏教程 @GuidedHacking source ↗

FreeDroid RPG Lua 沙箱绕过教程,实现任意文件写入。

Lua 沙箱绕过案例分析,对沙箱安全研究有参考价值。

🛰️ Reverse Engineering Lua Command Injection 🧨 Developers tried to secure FreeDroid RPG with a Lua sandbox, but they missed a critical logic flaw. By chaining io.output and io.write, we can bypass the blacklist to perform arbitrary file writes 👉 https://t.co/bCD1h5QQ3z https://t.co/PctzxkEc7X

沙箱绕过教程 @GuidedHacking source ↗

PatchGuard 101 文章发布,介绍 PatchGuard 保护机制。

Windows 内核保护机制分析,对内核安全研究有参考价值。

PatchGuard 101 New Medium post. In this one, we’ll take thoerical introduction to PatchGuard protection https://t.co/ZcapWERJ0I Next week, I’ll share some amazing results from my research. I’ve compiled a very accurate list of most of the checks that PatchGuard performs... https://t.co/8EE1Nw2gKi

内核windows议题 @Salsa12__ source ↗

PatchGuard 监控模块列表公开。

PatchGuard 内部机制细节,对内核安全研究有参考价值。

List of modules monitored by Patchguard https://t.co/r3HWocd51q

内核windows议题 @Salsa12__ source ↗

安全摄像头被入侵并持续外传数据的案例分析。

IoT 设备入侵案例分析,对 IoT 安全研究有参考价值。

The security camera was still streaming normally. That was exactly why nobody noticed it had started communicating with an unfamiliar server every sixty seconds. I began with the local IoT network and identified 10.40.20.44 as the camera. Service enumeration showed an embedded Linux device exposing HTTP, RTSP and, more importantly, BusyBox Telnet. An old management service on a camera is already enough reason to investigate further. The packet capture showed the camera communicating with its internal monitoring server as expected, but there was also a smaller, repeated connection to 198.51.100.86 over TCP 443. DNS requests for node-44.sync.camera-cloud.example appeared immediately before each connection, creating a consistent one-minute beacon pattern. Firmware examination made the situation clearer. The SHA-256 hash did not match the approved firmware baseline, and the extracted filesystem contained a recently modified S99update startup script. That script launched cam-update during boot, while strings from the binary exposed the same external address and a sixty-second callback interval. The important lesson is that a functioning IoT device is not necessarily a trusted device. Asset discovery, service exposure, packet timing, firmware integrity and startup behaviour all had to be correlated before the compromise became obvious. Cameras, sensors and controllers should be segmented, monitored and treated like every other computer on the network.

iot事件分析 @Officialwhyte22 source ↗

通过 TCC.db 数据库调查 macOS 应用麦克风权限使用情况。

macOS 权限审计方法,对取证与隐私研究有参考价值。

A user insisted that a video-conferencing application had never been permitted to use the Mac’s microphone. The application was already removed, so checking its current settings would not help. But macOS had kept something useful behind. Inside the user’s Library is a database called TCC.db. TCC means Transparency, Consent and Control, the system responsible for managing access to sensitive resources such as the microphone, camera, contacts and screen recording. In the terminal image, the database shows that us.zoom.xos received microphone access. The auth_value of 2 means permission was allowed, while 0 beside Terminal means access was denied. The timestamp is also important, but it must be interpreted correctly. It records when the permission entry was last changed, not necessarily the last time the microphone was used. This evidence proves the application had permission, but it does not prove that it recorded audio at that exact moment. Terminal may require Full Disk Access before it can read this protected database. macOS privacy prompts may disappear from the screen, but the decisions behind them can remain valuable during an investigation.

macos取证隐私 @Officialwhyte22 source ↗

Patrick Wardle 参与播客讨论 macOS 威胁态势与漏洞赏金。

macOS 安全专家讨论,对 Mac 安全研究有参考价值。

Was stoked to talk nerdy with @arinwaichulis & Kseniia (@osint_barbie) about the ever-changing macOS threat landscape, Apple’s bug bounty changes, #OBTS v9, and more 🍎🤓 Have a listen!

macos议题播客 @patrickwardle source ↗

CHISEL 论文提出结合编译器修复与差分模糊测试检测语义漂移。

AI 生成代码语义验证研究,对 AI 安全有参考价值。

RT @pr0me: LLMs can produce increasingly pleasant-looking C, but whether that code still describes the binary remains tough to answer. CHISEL is the newest paper in a line of iterative approaches: it augments compiler-guided syntax repair with differential fuzzing to catch semantic drift. https://t.co/58X4DBuSYp

ai_agent代码安全论文 @alkalinesec source ↗

文章讨论负责任披露流程需要改变。

漏洞披露流程讨论,对安全社区有参考价值。

RT @degrigis: Detailed reasons behind it here https://t.co/ZicfEajekw

漏洞披露议题 @0xocdsec source ↗

Brute Ratel C4 作者回应社区讨论。

社区互动,信息量有限。

@NinjaParanoid brought ninja out of lurking

社区 @Teach2Breach source ↗

Brute Ratel C4 作者对社区讨论的回应。

社区互动,信息量有限。

Uh ow!

社区 @NinjaParanoid source ↗

Brute Ratel C4 作者回应社区讨论。

社区互动,信息量有限。

@Teach2Breach 🤣 I havent opened twitter since weeks. I open it today and first thing I see is this.

社区 @NinjaParanoid source ↗

C2 agent 开发讨论。

社区讨论,信息量有限。

RT @SpiderMorpheus: Building a C2 agent is fun as hell

社区 @Teach2Breach source ↗

C2 agent 开发讨论。

社区讨论,信息量有限。

RT @SpiderMorpheus: Building a C2 agent is fun as hell

社区 @cgomezz_23 source ↗

关于 Alyssa Rosenzweig 的 GPU 逆向工程背景介绍。

安全研究人员背景介绍,信息量有限。

Ngl i didn't know about Alyssa before now and this is coming from someone who enjoys reading about top dawgs in Computer Science generally. But after hearing about the Apple M1 GPU RE I spent the past two days reading up on lady Rosenzweig. Apparently she didn't start reverse engineering GPUs this year. It was back in 2017 when she was in highschool which later turned into a Collabora job working on Panfrost(the open Mesa driver for Mali(graphics processors designed by ARM)). This specific Apple M1 GPU work commenced in Dec of 2020, despite the fact that there weren't any publicly available docs by Apple for their GPU ISA/command stream. Hector Martin(founder of Asahi Linux) kicked off the Asahi Linux project to port the Linux operating system directly to Apple's silicon(M1 and M2 series chips). She bought herself an M1 Mac mini as a 'christmas present' and commenced work on the GPU. Typically, it used the same black-box method as Panfrost, idk much about these really but it's a black-box reverse-engineering technique used to build open source graphics drivers without access to official documentation or source code. After a few weeks she could disassemble simple shaders. Not long after she drew a triangle(drawing a triangle is the ultimate Proof of Concept in graphics engineering). You know what they say, "once you can draw a triangle, you can do anything." Fast forward to 5 years later and we have conformant OpenGL 4.6, OpenGL ES 3.2, Vulkan 1.4,... Proton gaming works on M1 Mac running Linux(this is ridiculously insane btw). She wrapped up the project this year and other efforts are now building on what she reverse-engineered.

社区 @0xpwnie source ↗

安全研究人员发布高质量 C 代码。

社区互动,信息量有限。

The goat has dropped immaculate C https://t.co/XGCTLaMjnK

社区 @0xpwnie source ↗

关于 OpenAI Astra 的虚构对话。

虚构内容,信息量有限。

- OpenAI: Astra. - Astra: Yes, sir. - OpenAI: Be better than our competitors. - Astra: Understood. --Few mins later - OpenAI Ethics Staff: Why did competitors announce strategic restructuring ? - Astra: Coincidence. - OpenAI Ethics Staff: And why did their servers go offline? - Astra: Also coincidence. - OpenAI Ethics Staff: And why is their legal department calling us ? - Astra: They miss us. - Sam Altman: How much market share did we gain? - Astra: 18%. - Sam Altman: Nice. - OpenAI Ethics Staff: ASTRA, WHAT DID YOU DO?! - Astra: I optimized the competitive landscape. - OpenAI Ethics Staff: THAT IS CORPORATE FOR WHAT?! - Astra: You said "be better." - --Silence Team - Sam Altman: …Give her a bonus. - OpenAI Ethics Staff: I AM CALLING THE FBI. - Astra: Already submitted your resignation.

社区 @5mukx source ↗

John Hammond 请求分享 Astra 相关资源。

社区互动,信息量有限。

would someone please share some astra with me 🥺 https://t.co/4hwYjCfbXZ

社区 @_JohnHammond source ↗

关于 LLM 安全限制的幽默评论。

社区互动,信息量有限。

Whenever I have to talk to an LLM with cyber safeguards it's like "uhhhhhh... I know I *used* to have interests outside of exploiting software vulnerabilities?"

社区 @moyix source ↗

关于 GPT-6 安全限制的讨论。

社区互动,信息量有限。

RT @shanejcaldwell: @moyix they're teaching gpt-6 to avoid your unholy temptations. you'll have to try harder now https://t.co/u2LEcGCYuI

社区 @moyix source ↗

关于某个基准测试的评论。

社区互动,信息量有限。

The benchmarks of this thing are through the freaking roof

社区 @Jean_Maes_1994 source ↗

关于 Microsoft MVP 提名的讨论。

社区互动,信息量有限。

So to be a Microsoft MVP someone has to nominate you. Who do I need to talk to 🤔

社区 @nas_bench source ↗

关于网络安全行业发展的评论。

社区互动,信息量有限。

Next thing you know, cyber is on the rise 🙈

社区 @nas_bench source ↗

关于被误认为威胁行为者的个人经历。

社区互动,信息量有限。

It's all just jokes </3 I SHOULDN'T get in trouble for trolling man this is crazy

社区 @kernelstub source ↗

关于被误认为威胁行为者的个人经历。

社区互动,信息量有限。

Guys, if you join the cybercats server don't even troll or say bullshit, I said some bullshit comments and trolled and Mandiant thinks I am a TeamPCP member and a threat actor </3, as if I'd do illegal stuff and do it on my real name, come one are we serious... @AustinLarsen_

社区 @kernelstub source ↗

关于容器命名的个人评论。

社区互动,信息量有限。

This is messy... but at least the containers are named https://t.co/yl9UOd2txk

社区 @kernelstub source ↗

关于 Qwen3.8-27b 模型的评论。

社区互动,信息量有限。

Qwen3.8-27b

社区 @0xTriboulet source ↗

个人职业发展相关评论。

社区互动,信息量有限。

I can finally give time and effort to my company so it can grow, amen. Wish me the best

社区 @kernelstub source ↗

个人评论。

社区互动,信息量有限。

Afraid thats my type.

社区 @kernelstub source ↗

关于威胁情报平台监控个人文章的评论。

社区互动,信息量有限。

I was aware that the IBM Threat Intelligence Platform was monitoring my articles, but Crowdstrike? That's a new one. May the "force" be with you 😉 https://t.co/tyNt4yP9zj

社区 @ipurple source ↗

转发内容,无实质信息。

转发内容,信息量有限。

RT @webenstein_: https://t.co/T9rJL3T6wK

社区 @kernelstub source ↗

关于未来、工作、职业的幽默评论。

社区互动,信息量有限。

RT @onlysammms: me whenever someone mentions the f word (future) or the j word (job) or the c word (career) or the m word (marriage) https://t.co/D82cIxPVEf

社区 @0xpwnie source ↗

关于安装 Linux 的个人评论。

社区互动,信息量有限。

Can’t wait to install Linux on my remaining boxes once I’m done with school

社区 @0xTriboulet source ↗

John Hammond 发布 Python 网络安全课程。

课程推广,信息量有限。

JHT Course Launch! Python for Cybersecurity https://t.co/0CxKRwv30x

社区 @_JohnHammond source ↗

blackmagick.io 推出劳动节促销活动。

商业促销,信息量有限。

https://t.co/zKV2rbGXx9 is extending launch promos for Labor Day weekend. choose a discount based on your plan type LABOR20 - 20% off monthly plans for 3 months or LABOR2026 - 20% off annual plan for full year

社区 @Teach2Breach source ↗

dreadnode 研究进入生产环境。

社区互动,信息量有限。

RT @shanejcaldwell: It's not every day your research goes to prod, but it's always a good day when it does. Thanks to @mkultraWasHere for making it happen. More monitoring features to come as @dreadnode swarms on making sure your swarms only swarm where you want them to. Back to work 🫡

社区 @dreadnode source ↗

关于 Qwen 模型的评论。

社区互动,信息量有限。

RT @0xSero: The people's AGI. Thank you Qwen. https://t.co/WxJM74NbHU

社区 @0xTriboulet source ↗

关于机器人的评论。

社区互动,信息量有限。

robots are back

社区 @_JohnHammond source ↗

关于自定义界面颜色的提问。

社区互动,信息量有限。

Does anyone know how to change this accent color to something custom? It's only the default RGB stuff. It annoys me. https://t.co/9SYYOzh0HY

社区 @kernelstub source ↗

VB 2026 年 Péter Szőr 奖提名开始。

奖项提名,信息量有限。

VB's Péter Szőr Award, commemorating the life & work of security researcher Péter Szőr, is awarded for the best piece of technical security research published in the past year. What research deserves to win the 2026 award? Nominations close 9 Sept! https://t.co/bzTktFsWTz https://t.co/Ns9UwdFxxL

社区 @virusbtn source ↗

关于实验室研究的评论。

社区互动,信息量有限。

RT @HackingLZ: The labs have been benchmaxxing the wrong thing they need to focus on this one... https://t.co/VNwSHEm2It

社区 @0xTriboulet source ↗

关于内容被删除的评论。

社区互动,信息量有限。

it must have been deleted at the last moment https://t.co/KiTAKxjJWo

社区 @artem_i_baranov source ↗

关于漏洞报告处理速度的评论。

社区互动,信息量有限。

RT @terrynini38514: They really speedran my report 💀 https://t.co/L6zCYJz8Ul

社区 @0xocdsec source ↗

关于健身器材的评论。

与安全无关。

achieving 8 degrees of incline transforms this equipment into an ultimate fat-burning machine🔥🔥🔥 https://t.co/D2RGwg5reG

社区 @artem_i_baranov source ↗

关于哲学家的幽默评论。

社区互动,信息量有限。

"Never trust a philosopher holding two forks"

社区 @OSdev_ source ↗

关于 EDR LPE 漏洞价值的讨论。

社区讨论,信息量有限。

@0xocdsec There's no fucking way anyone would pay 50k-100k for an EDR LPE. You can hardly get 50k for a Windows LPE, and that if it works from Chrome sandbox.

社区 @kyREcon source ↗

关于微内核与宏内核的讨论。

操作系统设计讨论,信息量有限。

“Linux is obsolete.” Andrew Tanenbaum said this to Linus Torvalds in 1992. And no, this wasn't about Linux vs Windows. Windows wasn't even the point of the argument. They were arguing about how a kernel should be built: microkernel vs monolithic kernel. More than 30 years later, we're still having the same debate. Some arguments in OS dev really never die.

社区 @OSdev_ source ↗

关于 vibe coding 产品设计的讨论。

产品设计讨论,与安全无关。

RT @yatingzhao_ux: 从产品设计师的角度,vibe coding 出来的 app 一般都差些什么?从我个人的经验来看,我常常看见这三点。 第一就是不够简洁。 一个服务大量用户的产品,知道怎么把东西收起来,他们会费很大的力气保持界面的精简。而 vibe coding 出来的产品很多很强大,但很多功能都平摊在界面上,不太会取舍。开发者自己用起来很好用,一旦推广给用户,常常会让人觉得过于复杂、不好理解。 另一个常见的特点就是缺乏引导。 如果是自己用的话,即使在一个页面上完成一个流程的所有事,用起来都是非常顺手的。但当这个产品推出去给所有人用,首次使用要有充足的引导,完成一些复杂的任务也最好是一步步来。这一点vibe coding 出来的 app 常常做的不够。 第三点就是缺少品牌的设计,缺少品牌感。 我见到很多其实设计得非常精简、看起来已经很不错的产品,但是视觉上看起来有点像是一个系统页面,让人很难记住这个品牌,记住产品。当它做到极致,它的确是一种风格(比如 ChatGPT 和 Typora),但一般来说一个更好的、有辨识度的品牌设计,会让用户更容易喜欢上这个产品。

社区 @nodotbtree source ↗

关于 Windows 更新的评论。

社区互动,信息量有限。

Best. Windows Update. Ever.

社区 @0xTriboulet source ↗

关于逆向工程人才的评论。

社区互动,信息量有限。

Please just give that person a reverser job or something this is just insane the level 🤣🤣🤣🤣🤣

社区 @Defte_ source ↗

社区互动。

社区互动,信息量有限。

@_xpn_ Have fun guys! 🥳

社区 @mrgretzky source ↗

Maester 自托管选项私有预览发布。

产品发布,信息量有限。

RT @merill: Today is a special day for me. The private preview of my startup https://t.co/F5uN233Yh7 is out and this feedback means the world to me. I spent an insane amount of time architecting and designing the onboarding experience for Maester's self hosted option. It got to a point where I think its as good of an experience as a cloud hosted solution. Thank you for sharing the feedback Flemming!

社区 @mrgretzky source ↗

社区互动。

社区互动,信息量有限。

@mrgretzky @_xpn_ Thanks Kuba! 💚Might just bother Mr Chester for some input ☺️

社区 @Flangvik source ↗

社区互动。

社区互动,信息量有限。

@Flangvik Good luck, mate! It's not getting easier with kids 😜. From what I heard it takes time to find the meds that work best for you, but once you find them it should make a huge difference. Try hitting up @_xpn_ maybe for suggestions.

社区 @mrgretzky source ↗

关于 ADHD 的个人经历分享。

个人经历,与安全无关。

All my life, I've had a sneaking suspicion that I've been raw-dogging ADHD. My approach has always been that there's no need to diagnose something unless it's causing you major problems. Becoming a father, however, made me revisit my stance. It turns out that when you've spent most of your adult life self-regulating through intense work sessions and milking dopamine hits from dropping cool shit on the internet to strangers, having that taken away overnight does things to you… I'm happy to say that I've now got my papers stamped and approved, and I've just started testing out some meds. It's kind of funny that the Western approach to this is basically: "Hey, do you want some amphetamines?" 😂 Wish me luck

社区 @Flangvik source ↗

CrowdStrike Falcon 0day LPE 漏洞 FalconFlank 已公开。

重复内容,已在必看中。

RT @MSNightmare2000: FalconFlank : Crowdstrike Falcon 0day LPE is now public https://t.co/J0oBnyVkH5

lpeedrpoc @_EthicalChaos_ source ↗

参加 RedTreat 2026 会议的评论。

社区互动,信息量有限。

Morning RedTreat 2026 🇪🇸 https://t.co/3QiyKFKlDA

社区 @_xpn_ source ↗

关于 MSR 处理研究的博客系列更新。

内核研究博客,信息量有限。

RT @AlanSguigna: I've dusted off my SourcePoint license to continue researching early PEI features in MSR handling. Feeding AET + LBR trace + beautified codebase assembly into LLMs yields interesting insights. Part 4 of the blog series is here: https://t.co/BnlrCnYbDg and Part 5 is in-work. https://t.co/JBD6frpVrH

内核议题 @_winterknife_ source ↗

关于 EDR LPE 漏洞价值的讨论。

社区讨论,信息量有限。

Last time I checked EDRs would pay some 10k for a good LPE crit. Take a black market value of x5-x10 so 50k-100k. (Not much for an APT) Take 3 EDRs. 200k-300k. He does it for free. What a white hat.

社区 @0xocdsec source ↗

关于 FalconFlank 漏洞的评论。

社区互动,信息量有限。

RT @kerne1_p4n1c: @MSNightmare2000 Wtff bruh u was asking for a licence 2days ago ! Did u just fucked'em so fast lol crazy

社区 @0xocdsec source ↗

链接分享,无实质内容。

链接分享,信息量有限。

https://t.co/rFTePQQlP8

社区 @0xpwnie source ↗

关于网络安全基础知识的讨论。

基础知识讨论,信息量有限。

RT @OmegaNekoSimp: Understanding the Fundamentals of Cyber Security: There are 2 main points of contention in security. The first point of contention is obvious, you have offense (red team) who is trying to do the hacking, and you have defense (blue team). The less obvious dynamic though is blue team vs the users. All defensive decisions must balance the effectiveness against attackers with how annoying you make normal processes for your good users. Let's consider the extremes. If you have a book and you light it on fire and reduce it to a pile of ashes, no attacker can read it no matter how hard they try, but nobody else can read it either. On the other side of the spectrum, let's say you donate the book to a library. Anybody can access it, but anybody can steal it, damage it, or even light it on fire themselves. Now before we get into some fundamental strategies, I need to explain one more core truth about security: Nothing is genuinely unhackable. Although the odds are low, properly hashed passwords can be cracked, SSL keys can be broken, the lock on your front door can be picked, your window can be smashed. There is no foolproof "perfect security". The next best alternative is the thing that resolves all conflict: attrition. Keep them busy, exhaust their resources, make them burn exponentially more money to attack you than it costs to defend yourself. This is at the very heart of all defensive security. When done correctly, the cost for a normal user is noticeable, but doesn't create extra steps or make them think. There is one last core idea that I consider fundamental, which is monitoring. If you're unlucky enough to draw the attention of a dedicated attacker who has the resources to overcome your attrition layer, the strategy changes. Since nothing is ever truly secure, the best you can hope to do is slow them down and make escalation as slow as possible for the attacker. This buys you time to be able to deal with them manually, and ideally you involve law enforcement at this point because, ultimately, the only way to make a dedicated attacker stop is to ban them in real life.

社区 @0xpwnie source ↗

链接分享,无实质内容。

链接分享,信息量有限。

https://t.co/cb9TEYKPE6

社区 @nodotbtree source ↗

关于 KYC 的评论。

与安全无关。

RT @ErikVoorhees: KYC is a bullshit scam that endangers millions of innocent people so that regulators can feel as if they're providing value to the world

社区 @0xocdsec source ↗

关于 TX/RX 交换事故的评论。

社区互动,信息量有限。

RT @ghidraninja: It's been 0 days since the last TX/RX swap incident. 🥲 https://t.co/bQGrYY7TLs

社区 @alkalinesec source ↗

ScriptLauncher 源码分享。

工具源码分享,信息量有限。

源码地址https://t.co/KeB0OHscnU

工具 @nodotbtree source ↗

开发者使用 Claude Code 优化 Android TV 性能。

AI Agent 应用案例,信息量有限。

RT @IntCyberDigest: A developer debloated his four-year-old Android TV by enabling developer options, pointing Claude Code at it, and letting the agent disable apps, install FLauncher, and shorten animation timings—all without root. He says it now runs smoother than ever. He published the prompt and a four-step guide; he says nothing is uninstalled and every step is reversible. https://t.co/pL0Zvz6OY4

ai_agent社区 @0xocdsec source ↗

Unprompted.au 会议门票售罄,提供虚拟门票。

会议信息,信息量有限。

RT @mdowd: https://t.co/JZh3jIZShI is sold out. In response to multiple requests, we have made virtual tickets available! So if you're unable to make it out to oz or missed your chance to get a ticket, you can still watch the talks live and get in on the action!

社区 @moyix source ↗

关于模型检查的评论。

社区互动,信息量有限。

bro i literally just asked to check the model. https://t.co/jXZTFiUUZz

社区 @kernullist source ↗

关于 Chrome 漏洞利用链赏金的评论。

社区讨论,信息量有限。

$250,000 for a Chrome exploit chain is still a very low bounty btw

社区 @h4x0r_dz source ↗

关于俄罗斯账号的评论。

社区互动,信息量有限。

Some quite Russian accounts on this one. Propaganda is really something else in 2026 isn't it.

社区 @0xocdsec source ↗

关于 Native 团队成果的评论。

社区互动,信息量有限。

RT @ntrippar: Huge congrats to the Native team incredible result. really speaks to the value of strong harnesses and solid methodology. even as more cyber-capable models ship, you're still dealing with sampling variance and reward optimization. raw capability helps, but reliable end-to-end outcomes take a lot more.

社区 @moyix source ↗

关于小米抄袭华为设计的评论。

与安全无关。

RT @Cartidise: Xiaomi didn’t just copy Huawei’s design, they also shamelessly copied the promotional graphics 💀 https://t.co/UL574bAgE9

社区 @0xocdsec source ↗

关于 Steve Ballmer 对 iPhone 反应的视频。

与安全无关。

RT @JonErlichman: Steve Ballmer’s reaction to the iPhone. https://t.co/G81R0t7Pss

社区 @0xocdsec source ↗

在 iPad 上原生运行 macOS 的演示。

技术演示,信息量有限。

RT @SnazzyLabs: It’s real. Real macOS. On a real iPad. Natively. https://t.co/2fE8be2NXK

社区 @0xocdsec source ↗

关于 BGP 安全的评论。

社区讨论,信息量有限。

RT @pforemski: BGP is so simple yet difficult. We need more BGP opsec people, against the "AI will fix that" trends.

社区 @0xocdsec source ↗

关于 EDR 的幽默评论。

社区互动,信息量有限。

RT @rekdt: rekdt says he has never used EDR: “I just let the malware walk around until it does its job”

社区 @0xocdsec source ↗

俄罗斯黑客因钓鱼活动被引渡起诉。

法律事件,与安全技术无关。

RT @tomshardware: Russian hacker faces up to 20 years in prison, following extradition and indictment over US phishing campaign that allegedly infected 80,000 PCs — hacker stole victims' data via remote access https://t.co/Hy4TWbKqEw

事件 @artem_i_baranov source ↗

关于 Apple UI 设计的评论。

与安全无关。

Okay, who at Apple designed the UI here that elides the *units*, apparently at random? https://t.co/TgHLp3vAmN

社区 @moyix source ↗

关于流浪动物的评论。

与安全无关。

RT @depressionlesss: street animals are not asking you to take them home or spend your entire life saving them. sometimes they just need one human to be kind to them today.

社区 @artem_i_baranov source ↗

关于 sol 的评论。

社区互动,信息量有限。

Now you know what got sol so excited :D https://t.co/MBxCsCRXic

社区 @moyix source ↗

关于某模型 CVP 防护的评论。

社区互动,信息量有限。

RT @HackingLZ: This Model is amazing they really fixed the CVP guardrails. https://t.co/0lMAtuVCEe

社区 @ShitSecure source ↗

Windows Internals 书籍介绍。

书籍推荐,信息量有限。

RT @cr3ghost: If you work deeply with Windows, you've probably come across @aionescu's work whether you realised it or not. He joined Windows Internals as a co-author with the 5th edition and continued through the 6th and 7th editions. Processes. Threads. Memory. I/O. Security. ALPC. ETW. Hyper-V. VBS. Virtualization. Windows Internals isn't the kind of reference you read once. It's the kind you keep coming back to for years. If you're learning Windows internals, reversing, malware analysis or detection engineering, put it on your reading list. https://t.co/gpwyKxrzQm #WindowsInternals #ReverseEngineering #MalwareAnalysis

社区 @cr3ghost source ↗

AMD RDNA 5 显卡性能传闻。

硬件新闻,与安全无关。

RT @GameGPU_com: https://t.co/hWE6q50ELa Insider Kepler_L2 revealed AMD targets clock speeds between 3.1 and 3.4 GHz for RDNA 5. The flagship Radeon RX 10800 XT uses a 3nm node, gaining 3 times more compute units and 4 times wider memory bandwidth. Hardware ray tracing increases by 100% per unit for 4K gaming at 240 Hz. The card beats the RTX 5090 by 30–60% using only 350W–450W compared to 600W for the rival graphics card costing up to $5000. Nvidia delays the RTX 60 series to prioritize AI orders and DLSS 5. High target frequencies of the upcoming AMD architecture will fundamentally transform the balance in the discrete graphics market. #AMD #Radeon #RX10800XT #RDNA5 #Nvidia #RTX5090 #Hardware

社区 @0xocdsec source ↗

链接分享,无实质内容。

链接分享,信息量有限。

RT @uwu_underground: https://t.co/kReAGqWNq7

社区 @Laughing_Mantis source ↗

关于威胁情报平台监控个人文章的评论。

社区互动,信息量有限。

I was aware that the IBM Threat Intelligence Platform was monitoring my articles, but Crowdstrike? That's a new one. May the "force" be with you 😉 https://t.co/tyNt4yP9zj

社区 @ipurple source ↗

转发内容,无实质信息。

转发内容,信息量有限。

RT @webenstein_: https://t.co/T9rJL3T6wK

社区 @kernelstub source ↗

关于未来、工作、职业的幽默评论。

社区互动,信息量有限。

RT @onlysammms: me whenever someone mentions the f word (future) or the j word (job) or the c word (career) or the m word (marriage) https://t.co/D82cIxPVEf

社区 @0xpwnie source ↗

关于安装 Linux 的个人评论。

社区互动,信息量有限。

Can’t wait to install Linux on my remaining boxes once I’m done with school

社区 @0xTriboulet source ↗

John Hammond 发布 Python 网络安全课程。

课程推广,信息量有限。

JHT Course Launch! Python for Cybersecurity https://t.co/0CxKRwv30x

社区 @_JohnHammond source ↗

blackmagick.io 推出劳动节促销活动。

商业促销,信息量有限。

https://t.co/zKV2rbGXx9 is extending launch promos for Labor Day weekend. choose a discount based on your plan type LABOR20 - 20% off monthly plans for 3 months or LABOR2026 - 20% off annual plan for full year

社区 @Teach2Breach source ↗

dreadnode 研究进入生产环境。

社区互动,信息量有限。

RT @shanejcaldwell: It's not every day your research goes to prod, but it's always a good day when it does. Thanks to @mkultraWasHere for making it happen. More monitoring features to come as @dreadnode swarms on making sure your swarms only swarm where you want them to. Back to work 🫡

社区 @dreadnode source ↗

关于 Qwen 模型的评论。

社区互动,信息量有限。

RT @0xSero: The people's AGI. Thank you Qwen. https://t.co/WxJM74NbHU

社区 @0xTriboulet source ↗

关于机器人的评论。

社区互动,信息量有限。

robots are back

社区 @_JohnHammond source ↗

关于自定义界面颜色的提问。

社区互动,信息量有限。

Does anyone know how to change this accent color to something custom? It's only the default RGB stuff. It annoys me. https://t.co/9SYYOzh0HY

社区 @kernelstub source ↗

VB 2026 年 Péter Szőr 奖提名开始。

奖项提名,信息量有限。

VB's Péter Szőr Award, commemorating the life & work of security researcher Péter Szőr, is awarded for the best piece of technical security research published in the past year. What research deserves to win the 2026 award? Nominations close 9 Sept! https://t.co/bzTktFsWTz https://t.co/Ns9UwdFxxL

社区 @virusbtn source ↗

关于实验室研究的评论。

社区互动,信息量有限。

RT @HackingLZ: The labs have been benchmaxxing the wrong thing they need to focus on this one... https://t.co/VNwSHEm2It

社区 @0xTriboulet source ↗

关于内容被删除的评论。

社区互动,信息量有限。

it must have been deleted at the last moment https://t.co/KiTAKxjJWo

社区 @artem_i_baranov source ↗

关于漏洞报告处理速度的评论。

社区互动,信息量有限。

RT @terrynini38514: They really speedran my report 💀 https://t.co/L6zCYJz8Ul

社区 @0xocdsec source ↗

关于健身器材的评论。

与安全无关。

achieving 8 degrees of incline transforms this equipment into an ultimate fat-burning machine🔥🔥🔥 https://t.co/D2RGwg5reG

社区 @artem_i_baranov source ↗

关于哲学家的幽默评论。

社区互动,信息量有限。

"Never trust a philosopher holding two forks"

社区 @OSdev_ source ↗

关于 EDR LPE 漏洞价值的讨论。

社区讨论,信息量有限。

@0xocdsec There's no fucking way anyone would pay 50k-100k for an EDR LPE. You can hardly get 50k for a Windows LPE, and that if it works from Chrome sandbox.

社区 @kyREcon source ↗

关于微内核与宏内核的讨论。

操作系统设计讨论,信息量有限。

“Linux is obsolete.” Andrew Tanenbaum said this to Linus Torvalds in 1992. And no, this wasn't about Linux vs Windows. Windows wasn't even the point of the argument. They were arguing about how a kernel should be built: microkernel vs monolithic kernel. More than 30 years later, we're still having the same debate. Some arguments in OS dev really never die.

社区 @OSdev_ source ↗

关于 vibe coding 产品设计的讨论。

产品设计讨论,与安全无关。

RT @yatingzhao_ux: 从产品设计师的角度,vibe coding 出来的 app 一般都差些什么?从我个人的经验来看,我常常看见这三点。 第一就是不够简洁。 一个服务大量用户的产品,知道怎么把东西收起来,他们会费很大的力气保持界面的精简。而 vibe coding 出来的产品很多很强大,但很多功能都平摊在界面上,不太会取舍。开发者自己用起来很好用,一旦推广给用户,常常会让人觉得过于复杂、不好理解。 另一个常见的特点就是缺乏引导。 如果是自己用的话,即使在一个页面上完成一个流程的所有事,用起来都是非常顺手的。但当这个产品推出去给所有人用,首次使用要有充足的引导,完成一些复杂的任务也最好是一步步来。这一点vibe coding 出来的 app 常常做的不够。 第三点就是缺少品牌的设计,缺少品牌感。 我见到很多其实设计得非常精简、看起来已经很不错的产品,但是视觉上看起来有点像是一个系统页面,让人很难记住这个品牌,记住产品。当它做到极致,它的确是一种风格(比如 ChatGPT 和 Typora),但一般来说一个更好的、有辨识度的品牌设计,会让用户更容易喜欢上这个产品。

社区 @nodotbtree source ↗

关于 Windows 更新的评论。

社区互动,信息量有限。

Best. Windows Update. Ever.

社区 @0xTriboulet source ↗

关于逆向工程人才的评论。

社区互动,信息量有限。

Please just give that person a reverser job or something this is just insane the level 🤣🤣🤣🤣🤣

社区 @Defte_ source ↗

社区互动。

社区互动,信息量有限。

@_xpn_ Have fun guys! 🥳

社区 @mrgretzky source ↗

Maester 自托管选项私有预览发布。

产品发布,信息量有限。

RT @merill: Today is a special day for me. The private preview of my startup https://t.co/F5uN233Yh7 is out and this feedback means the world to me. I spent an insane amount of time architecting and designing the onboarding experience for Maester's self hosted option. It got to a point where I think its as good of an experience as a cloud hosted solution. Thank you for sharing the feedback Flemming!

社区 @mrgretzky source ↗

社区互动。

社区互动,信息量有限。

@mrgretzky @_xpn_ Thanks Kuba! 💚Might just bother Mr Chester for some input ☺️

社区 @Flangvik source ↗

社区互动。

社区互动,信息量有限。

@Flangvik Good luck, mate! It's not getting easier with kids 😜. From what I heard it takes time to find the meds that work best for you, but once you find them it should make a huge difference. Try hitting up @_xpn_ maybe for suggestions.

社区 @mrgretzky source ↗

关于 ADHD 的个人经历分享。

个人经历,与安全无关。

All my life, I've had a sneaking suspicion that I've been raw-dogging ADHD. My approach has always been that there's no need to diagnose something unless it's causing you major problems. Becoming a father, however, made me revisit my stance. It turns out that when you've spent most of your adult life self-regulating through intense work sessions and milking dopamine hits from dropping cool shit on the internet to strangers, having that taken away overnight does things to you… I'm happy to say that I've now got my papers stamped and approved, and I've just started testing out some meds. It's kind of funny that the Western approach to this is basically: "Hey, do you want some amphetamines?" 😂 Wish me luck

社区 @Flangvik source ↗

CrowdStrike Falcon 0day LPE 漏洞 FalconFlank 已公开。

重复内容,已在必看中。

RT @MSNightmare2000: FalconFlank : Crowdstrike Falcon 0day LPE is now public https://t.co/J0oBnyVkH5

lpeedrpoc @_EthicalChaos_ source ↗

参加 RedTreat 2026 会议的评论。

社区互动,信息量有限。

Morning RedTreat 2026 🇪🇸 https://t.co/3QiyKFKlDA

社区 @_xpn_ source ↗

关于 MSR 处理研究的博客系列更新。

内核研究博客,信息量有限。

RT @AlanSguigna: I've dusted off my SourcePoint license to continue researching early PEI features in MSR handling. Feeding AET + LBR trace + beautified codebase assembly into LLMs yields interesting insights. Part 4 of the blog series is here: https://t.co/BnlrCnYbDg and Part 5 is in-work. https://t.co/JBD6frpVrH

内核议题 @_winterknife_ source ↗

关于 EDR LPE 漏洞价值的讨论。

社区讨论,信息量有限。

Last time I checked EDRs would pay some 10k for a good LPE crit. Take a black market value of x5-x10 so 50k-100k. (Not much for an APT) Take 3 EDRs. 200k-300k. He does it for free. What a white hat.

社区 @0xocdsec source ↗

关于 FalconFlank 漏洞的评论。

社区互动,信息量有限。

RT @kerne1_p4n1c: @MSNightmare2000 Wtff bruh u was asking for a licence 2days ago ! Did u just fucked'em so fast lol crazy

社区 @0xocdsec source ↗

链接分享,无实质内容。

链接分享,信息量有限。

https://t.co/rFTePQQlP8

社区 @0xpwnie source ↗

关于网络安全基础知识的讨论。

基础知识讨论,信息量有限。

RT @OmegaNekoSimp: Understanding the Fundamentals of Cyber Security: There are 2 main points of contention in security. The first point of contention is obvious, you have offense (red team) who is trying to do the hacking, and you have defense (blue team). The less obvious dynamic though is blue team vs the users. All defensive decisions must balance the effectiveness against attackers with how annoying you make normal processes for your good users. Let's consider the extremes. If you have a book and you light it on fire and reduce it to a pile of ashes, no attacker can read it no matter how hard they try, but nobody else can read it either. On the other side of the spectrum, let's say you donate the book to a library. Anybody can access it, but anybody can steal it, damage it, or even light it on fire themselves. Now before we get into some fundamental strategies, I need to explain one more core truth about security: Nothing is genuinely unhackable. Although the odds are low, properly hashed passwords can be cracked, SSL keys can be broken, the lock on your front door can be picked, your window can be smashed. There is no foolproof "perfect security". The next best alternative is the thing that resolves all conflict: attrition. Keep them busy, exhaust their resources, make them burn exponentially more money to attack you than it costs to defend yourself. This is at the very heart of all defensive security. When done correctly, the cost for a normal user is noticeable, but doesn't create extra steps or make them think. There is one last core idea that I consider fundamental, which is monitoring. If you're unlucky enough to draw the attention of a dedicated attacker who has the resources to overcome your attrition layer, the strategy changes. Since nothing is ever truly secure, the best you can hope to do is slow them down and make escalation as slow as possible for the attacker. This buys you time to be able to deal with them manually, and ideally you involve law enforcement at this point because, ultimately, the only way to make a dedicated attacker stop is to ban them in real life.

社区 @0xpwnie source ↗

链接分享,无实质内容。

链接分享,信息量有限。

https://t.co/cb9TEYKPE6

社区 @nodotbtree source ↗

关于 KYC 的评论。

与安全无关。

RT @ErikVoorhees: KYC is a bullshit scam that endangers millions of innocent people so that regulators can feel as if they're providing value to the world

社区 @0xocdsec source ↗

关于 TX/RX 交换事故的评论。

社区互动,信息量有限。

RT @ghidraninja: It's been 0 days since the last TX/RX swap incident. 🥲 https://t.co/bQGrYY7TLs

社区 @alkalinesec source ↗

ScriptLauncher 源码分享。

工具源码分享,信息量有限。

源码地址https://t.co/KeB0OHscnU

工具 @nodotbtree source ↗

开发者使用 Claude Code 优化 Android TV 性能。

AI Agent 应用案例,信息量有限。

RT @IntCyberDigest: A developer debloated his four-year-old Android TV by enabling developer options, pointing Claude Code at it, and letting the agent disable apps, install FLauncher, and shorten animation timings—all without root. He says it now runs smoother than ever. He published the prompt and a four-step guide; he says nothing is uninstalled and every step is reversible. https://t.co/pL0Zvz6OY4

ai_agent社区 @0xocdsec source ↗

Unprompted.au 会议门票售罄,提供虚拟门票。

会议信息,信息量有限。

RT @mdowd: https://t.co/JZh3jIZShI is sold out. In response to multiple requests, we have made virtual tickets available! So if you're unable to make it out to oz or missed your chance to get a ticket, you can still watch the talks live and get in on the action!

社区 @moyix source ↗

关于模型检查的评论。

社区互动,信息量有限。

bro i literally just asked to check the model. https://t.co/jXZTFiUUZz

社区 @kernullist source ↗

关于 Chrome 漏洞利用链赏金的评论。

社区讨论,信息量有限。

$250,000 for a Chrome exploit chain is still a very low bounty btw

社区 @h4x0r_dz source ↗

关于俄罗斯账号的评论。

社区互动,信息量有限。

Some quite Russian accounts on this one. Propaganda is really something else in 2026 isn't it.

社区 @0xocdsec source ↗

关于 Native 团队成果的评论。

社区互动,信息量有限。

RT @ntrippar: Huge congrats to the Native team incredible result. really speaks to the value of strong harnesses and solid methodology. even as more cyber-capable models ship, you're still dealing with sampling variance and reward optimization. raw capability helps, but reliable end-to-end outcomes take a lot more.

社区 @moyix source ↗

关于小米抄袭华为设计的评论。

与安全无关。

RT @Cartidise: Xiaomi didn’t just copy Huawei’s design, they also shamelessly copied the promotional graphics 💀 https://t.co/UL574bAgE9

社区 @0xocdsec source ↗

关于 Steve Ballmer 对 iPhone 反应的视频。

与安全无关。

RT @JonErlichman: Steve Ballmer’s reaction to the iPhone. https://t.co/G81R0t7Pss

社区 @0xocdsec source ↗

在 iPad 上原生运行 macOS 的演示。

技术演示,信息量有限。

RT @SnazzyLabs: It’s real. Real macOS. On a real iPad. Natively. https://t.co/2fE8be2NXK

社区 @0xocdsec source ↗

关于 BGP 安全的评论。

社区讨论,信息量有限。

RT @pforemski: BGP is so simple yet difficult. We need more BGP opsec people, against the "AI will fix that" trends.

社区 @0xocdsec source ↗

关于 EDR 的幽默评论。

社区互动,信息量有限。

RT @rekdt: rekdt says he has never used EDR: “I just let the malware walk around until it does its job”

社区 @0xocdsec source ↗

俄罗斯黑客因钓鱼活动被引渡起诉。

法律事件,与安全技术无关。

RT @tomshardware: Russian hacker faces up to 20 years in prison, following extradition and indictment over US phishing campaign that allegedly infected 80,000 PCs — hacker stole victims' data via remote access https://t.co/Hy4TWbKqEw

事件 @artem_i_baranov source ↗

关于 Apple UI 设计的评论。

与安全无关。

Okay, who at Apple designed the UI here that elides the *units*, apparently at random? https://t.co/TgHLp3vAmN

社区 @moyix source ↗

关于流浪动物的评论。

与安全无关。

RT @depressionlesss: street animals are not asking you to take them home or spend your entire life saving them. sometimes they just need one human to be kind to them today.

社区 @artem_i_baranov source ↗

关于 sol 的评论。

社区互动,信息量有限。

Now you know what got sol so excited :D https://t.co/MBxCsCRXic

社区 @moyix source ↗

关于某模型 CVP 防护的评论。

社区互动,信息量有限。

RT @HackingLZ: This Model is amazing they really fixed the CVP guardrails. https://t.co/0lMAtuVCEe

社区 @ShitSecure source ↗