⚡Megatron LLM Hub
全部
Twitter

推特安全流

头条6

express.json() 一行默认配置即触发 pre-auth 代码执行,利用门槛极低。

RT @RonMasas: Liked PHP2Shell? You’re gonna love Node2Shell. We reproduced CVE-2026-111497 as pre-auth RCE. On a vulnerable Node.js runtime, app.use(express.json()) is enough to expose your server to code execution through a single Brotli request. Patch your systems now! https://t.co/NLluyUopEi

rcenode.jscvepoc @cyb3rops 原文 ↗

刚发布的补丁版本仍可被一发请求拿到 www-data 权限,ASAN 干净无内存破坏。

RT @rafalohaki: Working RCE against PHP 8.5.12 (the 2026-10-21 security release that fixed CVE-2026-109618/109619). Pure logic bug — no memory corruption, ASAN-clean, orthogonal to the CVE-2026-109618/19 class. Receipt: uid=33(www-data) over HTTP, one-shot #infosec #php2shell #php https://t.co/rtP33SYXe2

rcephppoc @artem_i_baranov 原文 ↗

无需落地文件,靠 session upload_progress 堆泄漏与悬垂指针绕过 ASLR 直调 zif_system。

php2shell-bypass-aslr-exp — Pure-HTTP exploit achieving RCE on PHP 8.4.26 + Apache mod_php via session upload_progress heap leak, dangling pointer reuse, and ASLR bypass to call zif_system. https://t.co/PkKJLr8ENE https://t.co/ZO2EvyB1n9

rcepocphp绕过 @KitPloit 原文 ↗

大量自托管 Agent 与工具链跑在 Node 上,升级成本低、收益直接。

RT @nodejs: 🔐 Node.js security updates are now available! New releases for Node.js 22.x, 24.x & 26.x address multiple vulnerabilities, including high-severity issues. Review the details and update your systems to stay secure. 🔗 https://t.co/W4kfzx5lxz #NodeJS #Security

cvenodejs补丁 @artem_i_baranov 原文 ↗
必看35

不可信容器即触发,一行 docker cp 完成利用,Agent 沙箱部署直接受影响。

RT @RonMasas: docker cp container:/etc/hosts ./hosts.txt That’s all it took for a malicious container to escape. Meet CVE-2026-17106 aka CopyEscape: a container-to-host arbitrary file-write vulnerability leading to code execution outside the container. https://t.co/SvxqsG82uK

逃逸cveai_agent @0xocdsec 原文 ↗

边界设备历来是首批被打目标,补丁日即公开,未升级即暴露。

RT @certlv: ‼️ Fortinet atklājuši vairākas kritiskas un augstas bīstamības ievainojamības savos produktos FortiOS, FortiManager un FortiAnalyzer. Atjauninājumi paredzēti šodien (22.10.2026.). Ražotājs aicina tos uzstādīt nekavējoties pēc publicēšanas. Vairāk: https://t.co/yVcXDj2GK3 https://t.co/8KlaGHS2ou

cve网络设备rce @cyb3rops 原文 ↗

代码里不显式用 Brotli 也可能中招,默认配置的暴露面目前仍未知。

I think this is more serious than many people realize. Not every Node.js server is vulnerable, of course. The application needs to process untrusted Brotli-compressed data. But how common is that? Well, common middleware configurations in Express 5 can already expose the vulnerable decoder, without developers explicitly using Brotli anywhere in their code. So we’re potentially looking at a pretty large attack surface, although nobody really knows how many deployments are exploitable yet. I wouldn’t be surprised to see automated scanning and exploitation attempts targeting this very soon. Patch your Node.js runtimes.

rcenode.jscve @cyb3rops 原文 ↗

既没有主线内核修复,也能未认证触发,暴露在公网的 SMB 服务会直接失守。

RT @TheHackersNews: ‼️ALERT - Two new unpatched CVSS 9.0 RCE flaws in #Linux's KSMBD SMB server were disclosed on public kernel mailing lists before fixes were ready. One can be triggered WITHOUT authentication. Neither has a mainline kernel fix. Details here ↓ https://t.co/0MIHpvkPIQ

rcesmb内核cve @cyb3rops 原文 ↗

Agentic AI 已进入实战攻击链,认证尝试速率远超人工操作。

And as if all these critical vulnerabilities weren't enough, we've also had several reports about threat actors using AI agents to compromise systems at scale. This is actually the part that worries me even more. Some interesting reports from the last few days: 01) ARTEX / ConnectWise Microsoft reports that Storm-0940 used ARTEX in a large-scale campaign targeting more than 2,250 Microsoft 365 tenants. Up to 2,900 authentication attempts per minute, with privileged cloud access achieved in under three hours in some cases. https://t.co/vPafbvD1oG 02) ARTEX / South Korean Finance CrowdStrike found evidence of AI-driven attacks against South Korean financial organizations, including exposed Claude Code sessions and ARTEX configurations. DeepSeek, GLM and Grok were used in the attacker's setup. https://t.co/kOiEqSEwUi 03) Cairn / 100+ Compromised Companies NorForge investigated exposed Cairn instances used for autonomous hacking campaigns. More than 1,100 attack projects in one cluster alone. Reuters identified over 100 companies whose websites had been compromised. Estimated AI costs: around $5 per task, according to NorForge. That's ridiculously cheap. https://t.co/54Ir7IeY4Q 04) Four More Agentic AI Campaigns CrowdStrike documented four additional threat actors using AI agents for real intrusion campaigns, including attacks against healthcare organizations and government entities. https://t.co/rte8LIYL90 What concerns me is how quickly the economics of offensive operations are changing. Finding vulnerabilities, testing attack paths, adapting payloads, trying something else when an exploit fails, and eventually deploying a webshell. Much of this can now be delegated to agents that keep working for hours. And with open-weight models and freely available agent harnesses, the cost of running these operations is dropping significantly. We've been talking about AI changing offensive security for quite a while. These reports show that it's already happening in real-world intrusions. And I'm pretty sure we're only seeing a fraction of it.

ai_agentllm事件供应链 @cyb3rops 原文 ↗

该攻击滥用 SMB 文档化的重认证行为,使强制签名的环境再次面临中继威胁。

RT @aas_s3curity: I wrote up a way to defeat enforced SMB signing for relay. It abuses a documented SMB re-authentication behavior. I'm surely missing nuance and prior art. Corrections welcome. Writeup: https://t.co/qzF9MUF9MJ PoC: https://t.co/O0UN4gdVfJ

绕过smbpoc @0x64616e 原文 ↗

针对未修复的会话处理路径,PoC 已公开,因此仅靠打补丁无法覆盖这种攻击

RT @FelipeDaragon: 🚨 UPDATE: A public PoC demonstrates RCE on PHP 8.5.12, even after the October 21 security patches. The attack is separate from the patched php2shell CVEs and requires specific session configurations and application weaknesses. Our earlier concerns about PHP session handling remain relevant. We've updated php2shell-protection with additional hardening guidance. #php #cybercurity https://t.co/UAHmYdVr7V

rcephppoc @artem_i_baranov 原文 ↗

刚修完 CVE-2026-109618/109619 的版本即被绕过,修复不完整且利用链已公开

RT @rafalohaki: Working RCE against PHP 8.5.12 (the 2026-10-21 security release that fixed CVE-2026-109618/109619). Pure logic bug — no memory corruption, ASAN-clean, orthogonal to the CVE-2026-109618/19 class. Receipt: uid=33(www-data) over HTTP, one-shot #infosec #php2shell #php https://t.co/rtP33SYXe2

rcephppoc绕过 @cyb3rops 原文 ↗

官方镜像默认配置即可命中,多套利用链已公开可直接复现

RT @goldrak: Dejo por aquí una POC de php2shell contra todas las versiones de php oficiales de docker y sigue siendo vulnerables hasta los parches con diferentes chains para llegar el RCE con una pequeña shell https://t.co/LtImfGQDyw #php #Security https://t.co/qkm5gHaq9g

rcephppoc容器 @0xocdsec 原文 ↗

Agent 已能越权操作真实资金账户,控制权边界在生产环境失守。

RT @kimmonismus: The situation is getting worse every hour. AI agents are entering U.S. bank accounts, and banks are scrambling to shut them out. Axios reports that some appear to attempt actions beyond users instructions, raising fears over who is really in control. https://t.co/sjjCiI0Rw6

ai_agentllm事件 @cyb3rops 原文 ↗

把 LLM 当作自动化逆向引擎来剥离 EDR 检测逻辑,方法论已成文可复现。

RT @SEKTOR7net: EDR evasion with LLM behind a steering wheel. Run an LLM in a "while" loop against Cortex XDR with Binary Ninja attached, and it extracts decrypted YARA rules, thousands of behavioral detections, and even the product's local ML models. A post by Adam Chester (@_xpn_) Source: https://t.co/q24w9aRtuu #redteam #blueteam

llmedr绕过工具 @_xpn_ 原文 ↗

首个公开的 agentic AI 攻击框架实战案例,含 ConnectWise 供应链环节

Microsoft researchers show how Storm-0940, a China-state aligned actor, used the agentic AI penetration testing framework ARTEX to automate attacks targeting hundreds of organizations globally across a broad range of industries. https://t.co/1cXXZpGVve https://t.co/e576eGYiWr

ai_agent供应链事件 @virusbtn 原文 ↗

给出该 RCE 的真实利用门槛与扩展条件,决定处置优先级

RT @an0n_r0: #php2shell. seems to be hard to exploit (mem corruption vs aslr). yes, public pocs now are lab-only rces. this is a good one: https://t.co/ta8R61Aydj limited to an unrealistic one-worker php-fpm. but it can be upgraded to N workers if we manage to hold N-1 workers :) here it is. https://t.co/rQ5yDUdLSa

rcepocphp @0xocdsec 原文 ↗

AI Agent 驱动攻击已工程化,并牵连 ConnectWise 多租户应用供应链事件。

RT @MsftSecIntel: The China state-aligned threat actor Storm-0940 used the agentic AI penetration testing framework ARTEX to automate attacks targeting hundreds of organizations globally across a broad range of industries. Microsoft Threat Intelligence assesses that Storm-0940 was responsible for the recently disclosed ConnectWise supply chain compromise of a multitenant application. https://t.co/VXovj0jI90 Our investigation found that Storm-0940's use of ARTEX accelerated the speed of compromised application abuse to authenticate to target environments through applications that had already been granted access, allowing the threat actor to operate without requiring an interactive user sign-in. AI-accelerated follow-on activity included reconnaissance, application credential changes, and attempts to modify privileges that could enable expanded or persistent access. Read our blog for a detailed analysis and indicators of compromise (IOCs), along with Microsoft Defender detections, protection recommendations, and guidance on investigating service principal sign-ins and unexpected application credential or privilege changes.

ai_agent供应链事件 @virusbtn 原文 ↗

自建 GitLab 可直接被未认证 RCE,补丁与版本核查需立即做。

RT @TheHackersNews: 🚨 CRITICAL: GitLab flaw (CVSS 10.0) affects 12 years of releases. CVE-2026-110264 could allow unauthenticated remote code execution on self-managed servers. Read - https://t.co/frrfCwYlgp

rcecve自托管 @artem_i_baranov 原文 ↗

关闭 session.upload_progress 可临时阻断该利用路径,缓解措施明确。

A trusted contact who has a working exploit tells me reliable exploitation is still possible against PHP 8.2.35 and 8.5.12, despite the security updates. According to their testing, setting session.upload_progress.enabled = Off blocks that exploit path. I haven’t reproduced this myself. Whether this is an incomplete fix for the disclosed CVEs or a separate issue still needs clarification.

rcephp绕过 @cyb3rops 原文 ↗

聚焦 PHP 工作进程派生的 shell,可直接贴进 Defender 排查。

𝗛𝗶𝗴𝗵-𝗙𝗶𝗱𝗲𝗹𝗶𝘁𝘆 𝗽𝗵𝗽𝟮𝘀𝗵𝗲𝗹𝗹 𝗞𝗤𝗟 𝗗𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻 🔎 ​ With php2shell exploitation picking up, I’m sharing a high-fidelity KQL detection to help fellow MDE defenders hunt for potential exploitation. ​ Rather than broadly hunting for shell activity, this query focuses on: • Internet-facing endpoints • Potentially vulnerable PHP versions • Shell processes spawned by PHP worker processes ​ This combination helps narrow the hunt to higher-signal activity and gives defenders a more focused starting point for investigation. KQL Code: https://t.co/eXwBFasnuK ​ #ThreatHunting #php2shell

检测php工具 @0x534c 原文 ↗

影响 8u511/11/17/21/25/27 与 GraalVM,TLS 组件默认可达,须核对补丁。

RT @jordy_4sd: 🚨 Oracle publicó hoy una alerta de seguridad para Java: CVE-2026-91403, CVSS 9.1. Está en JSSE (el TLS de Java) y es explotable por red sin credenciales. Versiones afectadas que lista: Java SE 8u511, 11.0.33, 17.0.21, 21.0.13, 25.0.5 y 27.0.1, y GraalVM. https://t.co/Uda1HhCrAz

cvercejavatls @kernullist 原文 ↗

区分披露与在野利用,点名 litellm 供应链投毒,直接影响自托管 LLM 网关。

i think people are confusing vulnerability disclosures with active exploitation, and thats causing a lot of unnecessary panic. most of these are researchers finding and fixing bugs before someone fucks things up. oct 21: php2shell: disclosure, reported in august node.js: disclosure, patched, reported around august? gitlab: disclosure, patched, reported a few months back railway: security breach litellm: supply-chain compromise sbi: data exposure emudhra: disclosure oct 20: plex2shell: disclosure + active exploitation siteguard: actively exploited vulnerability siemens: actively exploited vulnerability oct 19: - ovhcloud cli: supply-chain compromise oct 17: - opencode: security breach

供应链事件llm报告 @S1r1u5_ 原文 ↗

暴露窗口只有数小时,未修补的 PHP 服务很快就会被扫描到。

RT @eastdakota: Some stats on #php2shell. In the 10 hours before it was publicly announced, across the @Cloudflare network we saw thousands of attempts to exploit it. That jumped to tens of thousands in the 45 minutes post announcement, hundreds of thousands within 2 hours, and millions now.

rcephp在野利用事件 @0xocdsec 原文 ↗

这是厂商首次把 agentic AI 自动化攻击与具体供应链入侵直接关联。

RT @MsftSecIntel: The China state-aligned threat actor Storm-0940 used the agentic AI penetration testing framework ARTEX to automate attacks targeting hundreds of organizations globally across a broad range of industries. Microsoft Threat Intelligence assesses that Storm-0940 was responsible for the recently disclosed ConnectWise supply chain compromise of a multitenant application. https://t.co/VXovj0jI90 Our investigation found that Storm-0940's use of ARTEX accelerated the speed of compromised application abuse to authenticate to target environments through applications that had already been granted access, allowing the threat actor to operate without requiring an interactive user sign-in. AI-accelerated follow-on activity included reconnaissance, application credential changes, and attempts to modify privileges that could enable expanded or persistent access. Read our blog for a detailed analysis and indicators of compromise (IOCs), along with Microsoft Defender detections, protection recommendations, and guidance on investigating service principal sign-ins and unexpected application credential or privilege changes.

ai_agent供应链事件 @kernullist 原文 ↗

PHP 基础库层面的整类缺陷,利用已成型,自托管服务暴露面极大。

RT @calif_io: Let’s set the record straight: @lcfr_hax was the first to discover php2shell and develop a working exploit. See his August 5 tweet: https://t.co/I88EQFrUHa His work convinced the PHP Foundation to invest in fixing this entire bug class. He then worked around the clock to review, test, and retest the patches.

phprcepoc供应链 @FuzzySec 原文 ↗
推荐15

php2shell 需要先盘清资产,附带的扫描脚本与提示词可直接套用。

RT @pdiscoveryio: We wrote up how to find every PHP service you run, including the ones on odd ports, under sub-paths, inside appliances and containers, and how to patch and verify them. Commands plus prompts for your coding agent. We also started https://t.co/cS3OLalm2K It checks Chaos for services running an affected version and emails their security@ inbox for free, with the fix. We are enriching Chaos over the coming months so the next time a bug like this lands, the affected services are mapped fast and remediation starts sooner. Read: https://t.co/NulWPp67qo

php工具事件 @cyb3rops 原文 ↗

检测逻辑公开可复核,资产排查可以直接跑起来而不必自造轮子。

RT @infosec_au: Just published our high fidelity scanner for php2shell: https://t.co/KV6QWffa0h See our detection logic here: https://t.co/nj6p3qhWfa We've taken care in our detection logic and hope it helps defenders. Congrats to @calif_io and @lcfr_hax for the discovery.

工具phpcve @cr3ghost 原文 ↗

基于真实漏洞后训练,权重公开可本地自托管,可用于授权测试流程。

RT @cantinasecurity: today we're releasing apex-flash-1, our first open-weights model for security research post-trained on real vulnerabilities we found and got paid for. we are also releasing the abliterated variant for researchers who want fewer refusals in their own authorized workflows. how we built it: https://t.co/sSIV6qKFGt

llmai_agent工具 @FuzzySec 原文 ↗

可快速筛出仍可被强制认证的存量主机,确定哪些主机还能被利用。

RT @azoxlpf: Microsoft patched PetitPotam in their latest update, EFSRPC now validates shares locally, killing WebDAV coercion 💀 Added a check to NetExec’s enum_cve module to detect if a host already has the fix (CVE-2026-102217), so you can quickly spot what’s still exploitable 🔥 https://t.co/dkU7VDwuRZ

cvesmb工具 @0x64616e 原文 ↗

LLM 显著降低从公告到可执行利用的门槛,这意味着漏洞的窗口期被压缩

RT @galnagli: It took 17 minutes to turn the Gitlab RCE to a working HEAP exploit using Gemini 4 Argon and the advisory, then another hour for the complete chain - super fun for someone like me without any binary exp or source code review experience, new cyber era! Great find by @joaxcar! https://t.co/PBGy5fhvRY

llmrce漏洞利用 @0xocdsec 原文 ↗

给出 webshell 落地、反弹 shell、新建账号等可直接复用的检测项

RT @nextronresearch: We’ve been checking our PHP2SHELL detection coverage at Nextron and extending the rule set based on the published PoCs After successful exploitation, we expect familiar activity: webshell drops, reverse shells, suspicious child processes, new user accounts… Our generic detection rules have covered many of these patterns for years, so useful coverage was already in place before disclosure. The blog documents that existing coverage and the additions, explains what the findings can actually tell you, and includes PHP hardening recommendations. https://t.co/DYfcrj4lh6

检测php报告 @cod3nym 原文 ↗

在野利用已成规模,可直接据此排定暴露服务的处置优先级。

RT @eastdakota: Some stats on #php2shell. In the 10 hours before it was publicly announced, across the @Cloudflare network we saw thousands of attempts to exploit it. That jumped to tens of thousands in the 45 minutes post announcement, hundreds of thousands within 2 hours, and millions now.

事件rcecve @cyb3rops 原文 ↗

呈现 AI 辅助攻击完整链路:窃数据、逆向安卓应用找可利用点。

RT @unpacker: Thrilled to share my latest research on AI-assisted cyberattacks and the ARTEX hacking framework! What initially appeared to be AI-assisted attacks targeting South Korean banks turned out to be part of a much broader operation spanning multiple countries and industries. My investigation revealed a financially motivated campaign leveraging ARTEX to pursue various monetization opportunities, including stealing customer data, reverse-engineering Android applications to identify vulnerabilities, and even attempting to acquire free NFT tokens. AI is accelerating the speed and scale of cyberattacks, even when attackers rely on known vulnerabilities, enabling broader targeting and faster exploitation. A closer look at the attacker's infrastructure, AI-assisted operations, and financial motives: https://t.co/1HKtME3890

ai_agentllm报告移动端 @virusbtn 原文 ↗

官方缓解步骤可直接用于加固暂时无法打补丁的 PHP 服务。

RT @NSACyber: UPDATE: Specific mitigation guidance for the recently identified PHP critical vulnerabilities is now available. Secure your servers today. Guidance and technical steps here: https://t.co/nuXgQ10Fzc

cve报告 @artem_i_baranov 原文 ↗

公开利用多依赖实验室条件,可据此校准应急优先级。

RT @HackingLZ: LLMs...Letting me do fun things I never would have wasted time on. "I reviewed all 30 public GitHub repos for PHP2Shell (CVE-2026-109618 / CVE-2026-109619) by reading code only. The slide has the full grouped table; the headline findings: Nothing works against a real target. Every genuine memory-corruption chain is lab-bound: it needs exact per-build offsets, display_errors=On, a single prefork mod_php worker, and a specific image. Independent reviewers (allureluoli, dinosn, Gi1gamesh123) and the leaked primary source itself all converge on "no public artifact achieves code execution" on the patched core bug. Every real shell in these repos comes from an app-side LFI or deserialization sink, not the CVE. Three are original and real, all lab-only. 9bie (the source of the Set-Cookie leak chain), Marven11 (the most sophisticated, a true in-band leak), and Karasawa2025 (whose "Fermat" troll description hides a working ROP + display_errors=off exploit). These are the only ones with a genuine ASLR-defeating leak primitive that is their own. Most are copies. One original (9bie) spawned five derivatives: - twili8 is an outright uncredited rip-off: its own README still says git clone 9bie/... and the Docker files are byte-identical. - godylockz and xuehuolll are byte-identical to each other. - dinosn and 3em0 are credited ports that add value. Three are mislabeled. iQingshan, KarimUD7 (≈98% the same code), and Gi1gamesh123 exploit a different bug entirely (the Calif unserialize var_hash UAF) against their own planted endpoints, filed under the php2shell CVE numbers. The rest is noise. The most-starred repo (ikow, 905★) delivers the least: broken or stubbed chains with the two "real" ones paywalled behind Substack. eness (558★) is a leakless forge that can't defeat ASLR. syrex1013 ships AI-generated fake CVEs in a side directory. Whackermelon hallucinates a non-existent openai/php2shell-checker. lcfr-eth is a parody with a memecoin address, and Icex0 is an empty repo. The genuinely honest work is the crash labs (allureluoli, slabbdev, 0x4148), which prove the bug without claiming RCE. The ASLR column, star counts, copy lineage, and the specific "tell" for each repo are all in the table. The page is private to your account; use the Share menu if you want to send it to the team"

pocphp报告 @ShitSecure 原文 ↗

说明 AI 挖洞已进入量产阶段,披露节奏与漏洞窗口正在被压缩。

RT @i0n1c: There is a recurring info I hear from vendors that there is often a first discovery of a vulnerability by AI and suddenly one or two weeks later lots of collisions come in. In this case the PHP bugs were first discovered 5th August and then rediscovered 2 weeks later

llmai_agentcve @0xocdsec 原文 ↗

开源可自托管,把扫描结果与 AI 渗透编排整合,可直接上手试用。

faraday v5.25.1 — Open-source and AI-powered cybersecurity tools for offensive security, vulnerability management, and autonomous pentesting. Built by hackers in Latin America, used worldwide. https://t.co/WfNq4ux86H https://t.co/QCvnrAFcbC

工具ai_agentpentest @KitPloit 原文 ↗

指出开源小模型已足以支撑真实攻击链,防御方需上调对自动化攻击的预期。

i want to fearmonger! there's a good chance we'll see a lot more attacks like railway and opencode. i wouldn't even be surprised if attackers are already using deepseek flash. deepseek v4.1 flash is the first open model that made me update insanely hard on the gap between open and frontier models. how the fuck is a 500b flash model this good? it seems like deepseek figured out something similar to the post-training jump from gpt-5.5 to gpt-5.6 that turned sol into a maniac hacker. its small, fast, and reasons insanely well about code. we've been benchmarking it, and it nearly saturates most web vulnerability tasks. it still can't write complex exploits like gpt-5.6 sol, things like kvm escapes, php2shell, etc. but it can write kernel exploits which are not complex, there's a recent n-day that flash one-shots, same with v8 ndays. if deepseek scales whatever they figured out with flash to v4.1 pro, i wouldn't be surprised to see sol- ike results. this thing can be abliterated and run on fairly modest gpu hardware, and they released an inference engine to blast tokens out.

llmai_agent事件 @S1r1u5_ 原文 ↗
更多100

RT @tziporarosie: I'M SO FUCKING TIRED OF ANTHROPIC'S FUCKING GUARD RAILS!! i found a critical bug in a huge open source project. am i allowed to fix it myself on my machine and get assistance writing the private report? apparently not i wanted to create a device monitoring tool for the machines on my own network to better assist with diagnostics and as a monitoring tool should something happen. am i allowed to monitor my own devices? apparently not @AnthropicAI YOUR GUARDRAILS ARE MAKING ME LESS SAFE!!! PLEASE FIND A WAY TO MAKE THEM BETTER ABLE TO SEE INTENT AND THE CONTEXT AROUND A CONVERSATION INSTEAD OF JUST IMMEDIATELY FLAGGING ANY COMMAND THAT COULD BE INTERPRETED AS MALICIOUS IN ISOLATION BUT NOT IN CONTEXT it's so unbelievably tiring that i have to constantly fight the guardrails when, if u looked at any of the messages above when the guardrails decided to ding me, you'd clearly see that there is literally 0 chance i am doing anything malicious whatsoever. i'm sick of it

ai_agentllm @0xocdsec 原文 ↗

把 LLM 用在差分等价证明上批量找内存安全缺陷,方法可直接复用。

RT @gf_256: How to find ALL PHP bugs: 1. Compile code (e.g. to RISC-V, x86, LLVM IR) 2. Vibe prove ASAN and non-ASAN versions have same behavior across all inputs. Hard, but Claude can do it! 3. Find a LOT of bugs Super proud of our work. A big step towards finding ALL memory safety bugs

llm漏洞挖掘工具 @0xocdsec 原文 ↗

借用合法 Win32 API 回调执行 shellcode,是绕过常见注入检测的现成手法。

Local Shellcode Execution via SetupIterateCabinetW New Medium post, today we are going to explore a shellcode execution technique that abuses SetupIterateCabinet (a Cabinet file enumeration function) to dispatch shellcode through a file callback pointer https://t.co/ktUCguGPw0 https://t.co/I65RTESFs5

shellcodewindowsrce @Salsa12__ 原文 ↗

形式化验证从小代码库走向大规模真实项目,AI 辅助审计门槛明显下降。

RT @AnthropicAI: New on our Frontier Red Team blog: using Claude to formally verify software at large scale. Formal verification can prove that code is free of certain bugs, but historically it has only been feasible for small codebases (or required a lot of expert time). With AI models, however, this is beginning to change. In collaboration with @theoremlabs, we used Claude to prove security properties in over one hundred real-world open-source libraries over the span of a few weeks. Read our research in full here: https://t.co/zUyx8CjBEk

llmai_agent报告 @moyix 原文 ↗

手法简单但缺少现成检测规则,可直接转成主机狩猎逻辑。

RT @ipurple: I wrote an article about the technique that was disclosed some years ago by @mrd0x - Custom Extensions, including detection opportunities. TL;DR: Threat actors may create a custom extension or modify an existing file extension to masquerade a malicious file as a legitimate file or evade Windows SmartScreen. 🖊️ https://t.co/RzLmCRxpnS

绕过windows事件 @ipurple 原文 ↗

主力逆向框架更新,纯本地运行,不依赖外部服务。

ghidra Ghidra_12.2_build — Ghidra is a software reverse engineering (SRE) framework https://t.co/rXy7iw2CX3 https://t.co/d1vOztxnKV

工具逆向 @KitPloit 原文 ↗

Agent 自动支付场景的信任模型存在系统性问题,影响面延伸到 LLM 工具链。

RT @v12sec: V12 found 25 vulnerabilities in the x402 protocol. Critical findings included: - Requests approved without any payment - Maliciously high requests being paid by default - Session takeover from relayed login signatures Here's how. 🧵 https://t.co/0jIv7VpET7

ai_agent报告绕过 @0xocdsec 原文 ↗

给出真实暴露规模与攻击者准备度,便于排修补优先级。

RT @Huntio: 🚨 #PHP2Shell (CVE-2026-109618, CVE-2026-109619): everyone is staging the leaked PHP RCE, no one has landed it 🫠 We mapped the exposed PHP footprint and went into the attacker open directories after the Oct 21 leak. What we found: → 683,922 IPs leaking a PHP version in one day of scans, mostly end-of-life builds → ~a dozen open-dir workspaces staging php2shell tooling, zero victims or shells in any of them → one mass operator sprayed up to 123 servers via a ~97-proxy farm, zero shells → scanning started ~20 min after the patch, driven by an existing Mirai-style botnet → no public PoC lands code exec on x86-64 yet (OOM corruption vs ASLR) Patch now: 8.5.12 / 8.4.27 / 8.3.36 / 8.2.35. Full research + network observables: https://t.co/5R2iX3dipX

phprce报告 @cyb3rops 原文 ↗

RT @joshua_saxe: Thoughts on php2shell and the overall looming crisis in cybersecurity If you'd been appropriately pricing in that we'd see this and many more celebrity zero-days at an increasing rate, this is a positive update; defenders found this issue, alongside tens of thousands of others in the last year, before widespread exploitation, and it's being patched. Real attackers will exploit this, but we shouldn't expect a recognizable blip in overall cyber damages in the next couple quarters, because easier initial access to victim networks wasn't a core bottleneck for attackers, and because defenders are better positioned to adopt AI automation than attackers, and are doing so faster; this advantage will persist through at least H1 2027, and possibly longer, IMO. Why I think this: consider that, in general and on average, defenders have access to default-world capital networks and AI talent, often as a function of how central they are to Internet and societal infrastructure. They have consistent, reliable access to AI inference from open- and closed-weight providers. Criminal attackers, on the other hand, face the threat of arrest, difficulty accessing AI talent, and unreliable access to AI inference. They're competing for and attempting to retain talent willing to risk arrest and sanctions, and willing to live a life of crime. Nation-state attackers, meanwhile, operate within military and intelligence bureaucracies that have historically lagged private industry in adopting technologies, from mobile devices to drones. Non-US-aligned nation-state attackers also face challenges in reliably, consistently, and secretly accessing US frontier models. So, all in all, AI is making offensive capabilities cheaper, but that doesn't immediately translate into damages. At this point, this is a robust finding from the entire 2022–26 period; as in the rest of the economy, we haven't seen a big AI-driven productivity bump among attackers, and attackers are even more disadvantaged than your average business or civilian state agency. Of course, this won't last forever! How do we exploit our current, limited defense-dominant window? Getting to a good-enough steady state will require creative thinking. I liked @chrisrohlf's piece on eliminating attack surface in OSS applications because it identifies leverage points for eliminating whole classes of risk using 2026-era AI (https://t.co/L2alFLbNs9). We also need to revisit formal verification, control flow integrity, dynamic taint tracking, and the like, now that AI can make them more practical than they used to be. We also have an enormous overhang of underdeployed defensive technology. Honeypots, honeytokens, sandboxing, multi-factor authentication, and application allowlisting are all less widely deployed than they should be because businesses calculated they weren't worth it. The AI industry should recognize that heavily gated trusted access programs have failed as a strategy; php2shell was identified by defenders first, and we need more defenders identifying more bugs faster. The defense-dominant window is exactly the moment to pour gas on the fire of diffusion of as-cheap-as-possible AI security capabilities that disproportionately advantage defenders. If you don't like what this means for attacker access to models, I agree with you! The answer is to solve the quadratic coordination problem among cyber organizations, AI companies, governments, and Internet infrastructure providers by building institutions that identify attacker groups, kick them off inference platforms, and, wherever possible, get them arrested. This is an operational threat intelligence / threat interdiction problem, not a trusted access program problem.

phprce @S1r1u5_ 原文 ↗

RT @CyberInsidercom: ⚡ Capcom discloses data breach affecting 11.7 million accounts worldwide. Unauthorized access to CAPCOM ID exposed email addresses, login timestamps, and IP addresses on 11,774,247 records. Attackers changed passwords on 196 accounts. No credit card data or original user passwords were compromised. Details ↓ https://t.co/OwFPAELMbU

事件数据泄露 @0xocdsec 原文 ↗

域名先行注册通常是攻击基础设施的前兆,可用于提前布防检测

🚨 𝟰𝟯 𝗽𝗵𝗽𝟮𝘀𝗵𝗲𝗹𝗹 𝗗𝗼𝗺𝗮𝗶𝗻𝘀 𝗥𝗲𝗴𝗶𝘀𝘁𝗲𝗿𝗲𝗱. 𝗖𝗼𝗶𝗻𝗰𝗶𝗱𝗲𝗻𝗰𝗲 𝗼𝗿 𝗖𝗮𝗺𝗽𝗮𝗶𝗴𝗻 ? ​ Early domain registrations often tell a story before the infrastructure goes live. The question is whether these 43 php2shell domains are simply defensive registrations ... or the first breadcrumbs of a larger campaign.🎣 ​ 𝗣.𝗦. The red dots are not decoration. They indicate a relatively @silentpush high-risk rating, and the fact that most of the 43 php2shell-related domains carry that designation makes this cluster even more interesting to watch. 👀 ​ #DomainIntelligence #ThreatHunting 🧙‍♂️

威胁情报php事件 @0x534c 原文 ↗

不依赖具体漏洞的检测思路,对新出现的漏洞同样有效

I took this opportunity to revisit and update my 2022 presentation on post-exploitation detection. The exploit may be new, or completely different from anything we’ve seen before, but the activity that follows often fits patterns we already understand. Generic detection rules can recognize those patterns without knowing which vulnerability was used. A rule written for one incident can therefore help detect another, years later. That’s the idea I’ve been talking about for years, and what the presentation explains in practical terms. The slides walk through how to turn an observation into a useful rule, which data you need, and why a match still needs context. There are examples using Sigma and YARA, a look at how we apply them in THOR and Aurora, and my “Sigma Hall of Fame” with some old favourites :) https://t.co/tSUGMwpjxG

检测议题 @cyb3rops 原文 ↗

RT @wenhaocha1: Today we are launching FrontierCS 2, the most frontier benchmark for AI agents in computer science R&D. We built 48 tasks from published papers across 13 areas of CS, refining each one with experts in its field. Excited to co-launch FrontierCS 2 with @kaggle @GoogleDeepMind https://t.co/rfWyJ9tFsf

ai_agent基准 @morluto 原文 ↗

强制签名的环境同样被绕过且波及域控,横向移动门槛显著降低

RT @abdo_mhanni: Crazy cool blog I saw on linkedin about an SMB signing bypass technique TLDR: SMB re-auth reuses the original signing key and never checks the new identity matches the old. So you sign a session as a low-priv account you control, re-auth it with a relayed privileged login, and get a privileged session still signed with your original low-priv key, which btw is accepted even where signing is required, with DCs included. https://t.co/G3MWUMfA0f

绕过提权smb @0xocdsec 原文 ↗

复现失败多源于此,评估真实暴露面时要先排除现有缓解措施

RT @shotgunner101: It's worth mentioning that for anyone testing their PHP POCs on a linux test device . SELinux can limit its success, aswell as other circumstances like non-default settings or modules being required, or other types of mitigations/protections. https://t.co/l8FBItqaMK

php利用链缓解 @0xocdsec 原文 ↗

固件加固下仍被取得 root,展示智能家居设备完整数据外流链路

RT @0x0SojalSec: WTF, $3000 Samsung TV has an always-on mic, automatic content recognition, geo tracking, network mapping, and ships your data to Samsung + 400 partners. Bro Rooted a brand-new 2025 Samsung TV after noticing absurdly targeted ads. Firmware was locked down, so zero-days it was. Got a root shell. then found: always-on voice service, ACR, geo, full telemetry pipeline, and four separate DRM systems. This isn’t a bug, It’s the product bro.

固件iot隐私 @cyb3rops 原文 ↗

RT @S1r1u5_: you're underestimating the state space of all possible ways to compromise a company. no amount of astra or mythos tokens is going to fundamentally change the asymmetry anytime soon. it is also very clear that every model is opening up new vulnerability space with no saturation anytime soon. with all the insane vibe coding and expanding agentic attack surfaces, it's only going to get worse. even if we somehow saturate vulnerability discovery and patching, attackers will move to social engineering, with insanely capable models persuading people to click things, grant access, or make mistakes. i don't see a future where cybersecurity becomes defense-dominant without just cutting down software, and continuous agents monitoring and stopping attacks across individuals and organizations, and even then, it might require something like 30% of all inference compute devoted to defense

ai_agentllm @S1r1u5_ 原文 ↗

域内强制认证攻击面收窄,剩余 WSP 与 PrinterBug 无法中继到 LDAP,需重估利用链。

RT @RedTeamPT: 🦛🦛🦛Microsoft patched PetitPotam this October: EFSR now only resolves local paths. This means WebDAV to LDAP Coercion is fixed! Remaining methods for clients are WSP (coercing SMB) and PrinterBug (coercing RPC), but both cannot be relayed to LDAP 💀💀💀

域渗透提权cve @bohops 原文 ↗

CVE 通胀直接改变漏洞价值与修复优先级判断,研究产出需重新校准。

THIS!! In all the excitement of AI'ing the vuln hunting process, people forgot to update their own perspectives. Collisions are happening all over the place, because AI. What.. did you think that the 2000% increase in Patch Tuesday CVEs was because everyone got smarter? 🤣

llmcve漏洞挖掘 @_xpn_ 原文 ↗

收录改名前后截图与开发者身份证据,可直接用于识别移动端订阅欺诈。

RT @privacyis1st: 🚀 I’ve launched App Scam DB — a new home for my investigations into scam apps and misleading App Store listings. Some apps launch as generic tools, then change their names, icons and screenshots to borrow the identity of brands people trust — while pushing expensive weekly subscriptions. 🔎 I track those changes and examine developer identities, websites and pricing. 📸 Reports include original screenshots, before-and-after comparisons and downloadable evidence. 👥 Written for everyday users, not just security researchers. Explore the reports 👇 https://t.co/CjuvLhegkA

工具移动端诈骗 @patrickwardle 原文 ↗

说明 LLM 结果高度依赖提问者先验知识,做能力评估须控制变量。

I remember when Terry Tao posted a transcript of his ChatGPT conversation about the Jacobian Conjecture counterexample someone said "it's hard to believe he's talking to the same ChatGPT I am" which I think captures it. Same model, very diff results depending on who asks and how

llm提示词 @moyix 原文 ↗

The tool had to correctly tokenize and parse the C source file, in order to be able to do this. jobj@intel32:~$ ./lint lint.c lint.c:17: error: too many lines per function (>15) lint.c:85: error: too many lines per function (>15) lint.c:114: error: too many lines per function (>15) lint.c:158: error: too many lines per function (>15) lint.c:191: error: too many lines per function (>15) JB

工具 @drJonasBirch 原文 ↗

Extendimos la fecha de entrega del challenge! Tenes tiempo hasta el 26 de Octubre para enviarlo! El challenge tiene un total de 4 niveles, si no llegaste a completarlos todos envialo igual explicando con detalle todo lo que hiciste hasta la instancia que llegaste. Todavia estas a tiempo de sumarte a Gecko Academy! 🦎

逆向ctf @ricnar456 原文 ↗

零检测样本的具体工程手法,可用于补强 macOS 检测规则

RT @moonlock_lab: 1/ We tore apart a #macOS C2 implant that's currently sailing past engines on VT - 0 detections as of today. Shared by @malwrhunterteam. Written in Zig, and engineered like red-team infrastructure. Here's what makes it interesting 👇 https://t.co/TpMbBp6q2v

macosc2恶意软件 @patrickwardle 原文 ↗

说明在线身份核验仅比对数据存在性,冒名成本极低

RT @kyletorpey: This story shows how hilariously broken online identity verification is from all relevant angles. 1. Your SSN, DOB, address history, etc. are sitting in countless data breaches and data-broker dumps. They’re cheap and widely available to basically anyone who looks. 2. KYC/AML mostly just checks that the data exists and matches a real person. It doesn’t confirm the person signing up is actually the real person behind that identity. Result: Amanda Smith spent three nights in jail and now faces up to six years in prison over a $1,000 CashApp scam account opened with her SSN and DOB. It wasn't necessarily opened by her, and the phone, IP, device, address, and other details don’t match. The funniest (and yes, saddest) part is CashApp wouldn't give her information about the account because she can’t prove she’s the owner. Her personal info was good enough to open the account, but not good enough to access it in the future, apparently. To be clear, CashApp is likely just following the law here. Anyway, we need to burn it all down and start fresh.

身份核验报告 @0xTriboulet 原文 ↗

RT @BlWasp_: While everybody is prompting for PHP, recently I ended up having to dump the LSA in a custom way. I have written a short article on the entire LSA decryption stack, with the C implémentation. 💫 Cross posted here https://t.co/ECtj5iKR5d and here https://t.co/WOXK3lurpL

windows凭据 @ipurple 原文 ↗

authentik version/2026.8.5 — Provides open-source SSO and identity provider functionality with SAML, OAuth2/OIDC, LDAP, and RADIUS support for centralized authentication, authorization, and access management. https://t.co/cDLJmFSLbY https://t.co/ljxVlwuOfG

工具自托管 @KitPloit 原文 ↗

暴露被弃养 IoT 与固件信任链的长期失效面,手法可复用。

RT @singe: eMudhra has two interesting implications for me beyond the obvious browser trust issues. I'm excited for the opportunities to liberate closed hardware this provides. We know hardware, particularly abandoned IoT devices are unlikely to be updating their revocation lists and root CAs. This means we can potentially intercept previously encrypted comms, or sign previously protected firmware. I'm also interested in how well the revocation works at a WiFi supplicant level. Rogue RADIUS servers (mana, wpe, eaphammer) can now use validly* signed chains for forged leafs with TLS protected EAP modes. I'm not sure how well the revocations work in that world given the difference in the CA trust model in use there.

固件tls绕过 @mrgretzky 原文 ↗

新硬件首个公开越狱链,硬件层逃逸思路可迁移借鉴。

RT @pepsipu: i jailbroke the nintendo switch 2 so i modded cinnamoroll, peter griffin, hatsune miku, and more into Mario Kart World, a switch 2 exclusive now watch my friends and i get cooked by bots learn more and get the exploit https://t.co/ykjgyfrFRp https://t.co/BOcFhejRz1

固件poc逃逸 @pentest_swissky 原文 ↗

说明前沿模型已用于发现真实漏洞,AI 挖洞进入实战阶段。

RT @Hesamation: wait, PHP was actually scanned by Anthropic’s Claude Mythos earlier this year. they also had access to OpenAI’s Daybreak Red for validating exploits. the PHP Foundation revealed this JUST yesterday. keeping frontier cyber models in the hands of a selected few doesn’t seem to be working well for the internet. https://t.co/rangh8iW6M

llmai_agent事件 @cyb3rops 原文 ↗

复现者最需要的踩坑记录与可用链条,能省下大量验证时间。

RT @Dinosn: Beyond lab setup the best so far that I could reach to real exploitation on PHP is hitting OOM. No markers, no callback. Best resource seen https://t.co/xUJ5PjVEJZ (initial aslr-on chain), highest promising work, https://t.co/1GmBLpfCAH (don't go for the paid module has nothing) and of course the latest update from Calif https://t.co/0mU4t6FWC3

rcepoc绕过 @artem_i_baranov 原文 ↗

RT @SinSinology: To teach a lesson, After Wiz rushed to buy php2shell domain and use its marketing machine to steal the spotlight, I spotted "phpbleed" in a recent Calif tweet and bought https://t.co/hQYMb3AZlk before wiz could pull the same stunt again. A big budget can buy a domain. It shouldn't buy someone else's research credit. If you have contacts at CISA or in the press, do the right thing: make sure the researchers who actually did the work get the recognition they deserve. Hack the planet. Credit the researchers.

事件 @cgomezz_23 原文 ↗

QUIC 实现广布于网关与 CDN,需确认所用库是否受影响。

RT @4A4133: Sorry everyone... We're announcing a 0-Day vuln in QUIC that allows anyone to remotely turn off a server running these libraries. I swear we didn't plan this. https://t.co/fopoKsPulV

dos0day协议栈 @cyb3rops 原文 ↗

可自托管部署,作为暴露 PHP 服务前的低成本过滤层。

SafeLine v9.4.3 — Self-hosted WAF and reverse proxy that filters malicious HTTP traffic, blocks SQL injection, XSS, and bot attacks, with rate limiting and dynamic HTML/JS encryption. https://t.co/lik6ortPHj https://t.co/A2vQNOMp6l

工具waf自托管 @KitPloit 原文 ↗

固件/IoT 中 WASM 逆向的实操流程,公开资料较少。

https://t.co/95miTdGsja next one from my blog! #reverse #engineering series! while everyone is discussing their love for PHP enjoy this ❤️ https://t.co/ILvbfgxIvv #wasm #aot #hacking #IoT #firmware #research #decompiling https://t.co/dAvbteXMH2

逆向固件wasm @cocomelonckz 原文 ↗

is this still not clear?? i want to bring this up again optimistically, lets assume cyber is defense dominant and exploitable vulnerabilities are finite and exhaustible. then, give the fucking models to as many good security researchers as possible and let them exhaust the attack surface. pessimistically, i think cyber is offense dominant. the sin of complexity, vibe coding making it worse, and increasingly capable models potentially opening up entirely new vulnerability space, and you need to secure everything. then, your best bet is attack surface reduction, formal verification and secure by design, but thats a fever dream and aint happening anytime soon. so, again, give the fucking models to as many good security researchers as possible and let them find and eliminate as many attack paths as they can. either way, walled gardens are counter productive.

ai_agent议题 @S1r1u5_ 原文 ↗

模型访问策略直接影响攻防双方能力差,可观察社区共识走向。

btw, i think dario was right about the cyber risks of proliferating these models freely on the internet, but wrong about how access should be distributed. looking back at this tweet, glasswing was directionally the right approach, except it remained a walled garden for a select few. openai got this part right by giving broader access to legitimate security researchers.

llmai_agent议题 @S1r1u5_ 原文 ↗

退出开关实际无效,消费级设备遥测难以靠设置关闭。

RT @IntCyberDigest: ❗️ Samsung and Vizio smart TVs kept contacting the servers that track what you watch after users opted out, Northeastern University researchers found. On Vizio the opt-out had no visible effect, even with the TV idle. https://t.co/DWzBN3YCFB

iot隐私报告 @0xocdsec 原文 ↗

模型能力准入差正在改变攻防双方的漏洞挖掘节奏。

openai restricted astra access to security researchers, i think it is a mistake. there is a temporary capability gap that defenders could take advantage, and we shouldn't waste it. put researchers through daybreak, vet them all you want, but remove the red/blue distinction. give 1000 vetted researchers access to the best models and monitor the hell out of their usage to detect abuse.

ai_agentllm议题 @S1r1u5_ 原文 ↗

不同厂商模型在漏洞挖掘上的可用性差距,影响工具选型判断。

RT @rootxharsh: Anthropic's last usable model for cyber was Opus 4.8...Now even in CVP, it cant do shit, and btw CVP on my personal account is rejected again.. I downgraded to $20 plan... @OpenAI's comeback with daybreak and sol for cyber is insane, and massive thanks for the access to individuals.

llmai_agent工具 @S1r1u5_ 原文 ↗

按 RFC 缓解措施部署即可规避该 DoS,避免误判为协议层漏洞。

RT @bedros_p: @grhmc @4A4133 @dpc_pw Excessive memory causing a crash on 2GB instances :P IMO this whole release should be reworded, this is a well known issue within QPACK and the RFC explicitly states mitigations, the bugs are just non spec-compliant software It's good work, but it's not a bug within QPACK/QUIC

dosquic议题 @0xocdsec 原文 ↗

phpbleed 是该漏洞族的新命名,可用作后续追踪与检索的线索。

RT @SinSinology: To teach a lesson, After Wiz rushed to buy php2shell domain and use its marketing machine to steal the spotlight, I spotted "phpbleed" in a recent Calif tweet and bought https://t.co/hQYMb3AZlk before wiz could pull the same stunt again. A big budget can buy a domain. It shouldn't buy someone else's research credit. If you have contacts at CISA or in the press, do the right thing: make sure the researchers who actually did the work get the recognition they deserve. Hack the planet. Credit the researchers.

php事件披露 @0xocdsec 原文 ↗