⚡Megatron LLM Hub
全部
Website

rss-website

头条1

网络栈默认路径受影响且无需认证,暴露公网的主机面临直接利用风险。

CVSSv3 Score: 9.8 CVE-2026-98432:ipv4: free inet_opt and ireq_opt after an RCU grace period tcp_v4_syn_recv_sock() transfers ownership of ireq->ireq_opt to the child socket (newinet->inet_opt) without copying it. Another cpu can concurrently process a retransmitted SYN for the same request socket, and send a SYNACK from tcp_check_req(). tcp_v4_send_synack() and inet_csk_route_req() read ireq->ireq_opt under rcu_read_lock() only, and ip_build_and_send_pkt() and ip_options_build() then read opt->optlen twice. Note that the SYNACK timer itself is not an issue: it holds a reference on its request socket, and inet_csk_reqsk_queue_drop() calls timer_delete_sync() before the child can be freed. Since commit 079096f103fa ("tcp/dccp: install syn_recv requests into ehash table"), request sockets are processed without holding the listener lock, so nothing prevents the child socket from being freed while the SYNACK is still being built. TCP child sockets do not have SOCK_RCU_FREE, and inet_sock_destruct() frees inet_opt with a plain kfree(), leading to a use-after-free in ip_options_build(). A similar issue exists with request socket migration (net.ipv4.tcp_migrate_req=1, or a BPF_SK_REUSEPORT_SELECT_OR_MIGRATE program). reqsk_timer_handler() clones the request socket with inet_reqsk_clone(), so that the old request socket and its clone share the same ireq_opt, then reqsk_migrate_reset() clears the pointer in the old one. Another cpu holding a reference on the old request socket can still be using these options (sending a SYNACK, or creating a child in tcp_v4_syn_recv_sock()) when the clone is freed, and tcp_v4_reqsk_destructor() also uses a plain kfree(). Readers already use RCU, and other paths replacing inet_opt (do_ip_setsockopt(), cipso_v4_sock_setattr()...) already use kfree_rcu(). Use kfree_rcu() in inet_sock_destruct() and tcp_v4_reqsk_destructor() as well. IPv6 is not affected by the first issue, because tcp_v6_syn_recv_sock() duplicates the options. tcp_v6_reqsk_destructor() has the same migration issue with ipv6_opt, which is only set by CALIPSO. This will be addressed in a separate patch. Revised on 2026-11-02 00:00:00

内核uafrcecve @forti-ir 原文 ↗
必看1

反序列化白名单这一缓解措施本身失效,加载不可信模型数据即执行代码。

Microsoft Qlib RestrictedUnpickler Deserialization of Untrusted Data Remote Code Execution Vulnerability November 2nd, 2026 ZDI-26-849 ZDI-CAN-32134 CVE ID CVSS Score 7.8 AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H Affected Vendors Microsoft Affected Products Qlib Vulnerability Details This vulnerability allows remote attackers to execute arbitrary code on affected installations of Microsoft Qlib. User interaction is required to exploit this vulnerability in that the target must visit a malicious page or open a malicious file. The specific flaw exists within the handling of pickle data files. The issue results from the lack of proper validation of user-supplied data, which can result in deserialization of untrusted data. An attacker can leverage this vulnerability to execute code in the context of the current user. Additional Details Fixed in v0.9.8 https://github.com/microsoft/qlib/releases Disclosure Timeline 2026-07-17 - Vulnerability reported to vendor 2026-11-02 - Coordinated public release of advisory 2026-11-02 - Advisory Updated Credit YJK(@YJK0805) of ZUSO ART Back to Advisories

rce反序列化绕过cve @zdi-advisories 原文 ↗
推荐5

量化研究流水线常加载外部记录文件,CVSS 7.8 且需一次用户交互。

Microsoft Qlib MLflowRecorder Deserialization of Untrusted Data Remote Code Execution Vulnerability November 2nd, 2026 ZDI-26-850 ZDI-CAN-32135 CVE ID CVSS Score 7.8 AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H Affected Vendors Microsoft Affected Products Qlib Vulnerability Details This vulnerability allows remote attackers to execute arbitrary code on affected installations of Microsoft Qlib. User interaction is required to exploit this vulnerability in that the target must visit a malicious page or open a malicious file. The specific flaw exists within the MLflowRecorder class. The issue results from the lack of proper validation of user-supplied data, which can result in deserialization of untrusted data. An attacker can leverage this vulnerability to execute code in the context of the current user. Additional Details Fixed in v0.9.8 https://github.com/microsoft/qlib/releases Disclosure Timeline 2026-07-17 - Vulnerability reported to vendor 2026-11-02 - Coordinated public release of advisory 2026-11-02 - Advisory Updated Credit YJK(@YJK0805) of ZUSO ART Back to Advisories

rce反序列化cve @zdi-advisories 原文 ↗

三条同批披露中的第三个攻击面,数据加载入口同样未做类型校验。

Microsoft Qlib StaticDataLoader Deserialization of Untrusted Data Remote Code Execution Vulnerability November 2nd, 2026 ZDI-26-848 ZDI-CAN-32133 CVE ID CVSS Score 7.8 AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H Affected Vendors Microsoft Affected Products Qlib Vulnerability Details This vulnerability allows remote attackers to execute arbitrary code on affected installations of Microsoft Qlib. User interaction is required to exploit this vulnerability in that the target must visit a malicious page or open a malicious file. The specific flaw exists within the StaticDataLoader class. The issue results from the lack of proper validation of user-supplied data, which can result in deserialization of untrusted data. An attacker can leverage this vulnerability to execute arbitrary code in the context of the current user. Additional Details Fixed in v0.9.8 https://github.com/microsoft/qlib/releases Disclosure Timeline 2026-07-22 - Vulnerability reported to vendor 2026-11-02 - Coordinated public release of advisory 2026-11-02 - Advisory Updated Credit Juan Pablo Lopez Yacubian Back to Advisories

rce反序列化cve @zdi-advisories 原文 ↗

补充受影响面判断与检测思路:Go 写的服务默认监听即可能中招。

A critical vulnerability in the Linux kernel’s IPv4 networking code, tracked as CVE-2026-98432 (CVSS 9.8), allows potential remote, unauthenticated attackers to trigger a use-after-free condition on exposed systems. This vulnerability specifically affects Multipath TCP (MPTCP) handling, but could also impact standard IPv4 traffic as well. Because Sandfly is built on Go, and recent Go toolchains (Go 1.24+) enable MPTCP on listening sockets by default, we have received questions regarding Sandfly's exposure. The vulnerability is not in Sandfly's code, but the Linux kernel itself and affects all modern Go infrastructure server ecosystems such as Kubernetes, Docker, and many others. We advise all teams to audit and update their Linux systems that run the Sandfly server (UI and API) and nodes as a precaution.

内核mptcp检测cve @sandflysecurity-blog 原文 ↗

说明补丁管理不足以拦截入侵,需对已知暴露面做持续验证。

Oct 29, 2026 Why Decade-Old CVEs Are Still Being Used to Breach Networks in 2026 CISA’s latest KEV additions show that patch management alone can’t catch what continuous exposure validation would. Summary On October 8, 2026, CISA added five vulnerabilities to its Known Exploited Vulnerabilities (KEV) catalog. None are new disclosures and some date back to 2015. Patches have existed for all five vulnerabilities for years, and attackers are still exploiting them today. In many cases, the likely reason is that organizations don’t know they’re still exposed. Point-in-time vulnerability scans can create a false sense of coverage. A vulnerability that looks patched may still be reachable, misconfigured, or reintroduced through rollbacks, shadow IT, M&A, or third-party assets. CISA’s new federal remediation directive (BOD 26-04) ties KEV status to firm deadlines, which raises the cost of not knowing what you’re exposed to. Continuous threat exposure management (CTEM) helps close that gap, prioritizing exploitability and validating defensive measures on an ongoing basis instead of waiting for the next audit cycle. Author: Francisco Najera, Sales Engineering Manager Continuous threat exposure management (CTEM) is a security program model that continuously scopes, discovers, prioritizes, validates, and remediates an organization’s real-world exposure, rather than relying on periodic point-in-time scans. CISA’s newest Known Exploited Vulnerabilities (KEV) additions are a good example of why that approach matters. On October 8, 2026, CISA added five vulnerabilities to the KEV catalog. None of them are new. The oldest, ISC BIND and ProFTPD, were first disclosed in 2015—that’s eleven years ago. The newest in this group is in Strapi and goes back to 2023. Patches have existed for all five for years. Even so, attackers are exploiting them now, and in production. This looks less like a patching failure and more like a visibility failure. Organizations don’t get breached through a vulnerability with a three-year-old patch because the fix doesn’t exist. It happens because the vulnerable system was never on anyone’s radar or because no one confirmed that the fix was correctly applied. This is exactly where CTEM helps. By testing and validating your compensating controls on a continuous basis, you can confirm that even if a system remains vulnerable, your defenses will disrupt the attack and limit the impact. Which Vulnerabilities Did CISA Add to the KEV Catalog in October 2026? CISA confirmed active exploitation for all five of the KEV entries: Strapi, CVE-2023-22894 : cleartext storage of sensitive information ONLYOFFICE Docs Server, CVE-2021-3199 : path traversal Apache Struts, CVE-2016-3081 : command injection ISC BIND, CVE-2015-5477 : data processing errors ProFTPD, CVE-2015-3306 : improper access control The CISA KEV catalog exists for a simple reason: to help organizations prioritize which vulnerabilities to focus on based on real world exploitability vs a theoretically rank CVSS score. That makes the list useful, and a little uncomfortable. Every entry is evidence that somewhere, a patchable system stayed exposed long enough for an attacker to find it first. How Does CISA’s BOD 26-04 Change Vulnerability Remediation? This KEV update lands under a new federal remediation regime. CISA’s Binding Operational Directive 26-04 , issued in June 2026, replaces flat “patch within X days” timelines with risk-based deadlines. KEV status is one of the factors, along with internet exposure and how much control an attacker gains. Federal agencies feel that pressure directly, and for everyone else, it signals where auditors and regulators may be heading. A known, exploited, unpatched vulnerability is more than a security gap. It can also become a compliance problem. Why Do Patched Vulnerabilities Keep Getting Exploited? It’s tempting to read this KEV update as purely a patch management story. It’s not, or at least not only. Older CVEs with available fixes survive this long partly due to the sheer volume of vulnerabilities, because teams are slow to close tickets, and partly because of exposures nobody is tracking: Shadow IT and forgotten assets. A dev team spins up a Strapi instance for a project or may use collaboration tools without knowing ONLYOFFICE components are leveraged for a project. The project ends, but the systems remain online. M&A-inherited infrastructure. Acquired environments often carry years of unpatched legacy software that isn’t in the parent company’s scope on day one. Configuration drift. A patch gets applied, then a rebuild, rollback, or misconfigured image quietly brings the vulnerable version back. Third-party and vendor exposure. The vulnerable instance isn’t yours. It belongs to a partner or supplier with access to your environment. A vulnerability scan run once a quarter can’t see any of this in real time. It only tells you what was true when the scan ran, for the assets it knew to look for. How Does CTEM Close the Exposure Validation Gap? This is the gap continuous threat exposure management is meant to close. Instead of treating exposure as a quarterly checklist, CTEM treats it as an ongoing cycle: scope the attack surface, discover exposures, prioritize by real-world risk, validate whether they’re actually exploitable, and mobilize teams to fix what matters most. Validation is the step many programs skip, and it’s the step that provides the strongest defense despite the systems not being previously patched. Validation doesn’t ask, “Is this CVE present somewhere in our environment?” It asks, “Can an attacker actually reach and exploit this, and where can they go next?” That’s the difference between a vulnerability report and a prioritized, provable list of what’s really at risk today. This is where adversarial exposure validation comes in. It continuously and safely simulates the techniques attackers use against known exploited CVEs like these, so security teams get proof instead of an assumption that a given exposure is, or isn’t, a real path into the environment. What Should Security Leaders Do About These KEV Additions Right Now? Search your asset inventory by product name. Lookout for Shadow IT and demand the same from third parties. Search for Strapi, ONLYOFFICE Docs Server, Apache Struts, ISC BIND, and ProFTPD, not just the CVE IDs. Verify, don’t assume. Confirm the patch was actually applied on the running instance, not just closed in a ticketing system. Prioritize internet-facing systems first. That’s where CISA and attackers are both looking. Shift from periodic scanning to continuous validation. A scan tells you a vulnerability exists. Validation tells you whether it’s actually exploitable in your environment. To see how the SafeBreach CTEM platform continuously validates exposure against known exploited vulnerabilities, including the ones already on CISA’s KEV catalog, across your real environment, schedule a demo . About the Author Francisco Najera , is a Sales Engineering Manager at SafeBreach, the leader in enterprise-grade exposure validation. With over 25 years of technical experience, Francisco works closely with enterprise leaders across major industries, helping their teams validate their defenses and mature their CTEM programs. Frequently Asked Questions What is the CISA Known Exploited Vulnerabilities (KEV) catalog? The CISA Known Exploited Vulnerabilities (KEV) catalog is a list of CVEs that CISA has confirmed are being actively exploited in real-world attacks, not just rated as severe. Federal civilian agencies must remediate KEV entries by set deadlines. Many private-sector security teams also use it to prioritize, since every entry reflects confirmed, in-the-wild exploitation rather than theoretical risk. Why do attackers still exploit vulnerabilities that are years old? A patch existing doesn’t mean every instance got patched. Shadow IT, forgotten test servers, M&A-inherited systems, configuration drift, and third-party vendor assets can all keep long-patched CVEs reachable. Old exploits also tend to be well documented and reliable, which makes them cheap for attackers to reuse. What’s the difference between vulnerability management and exposure validation? A vulnerability scan reports whether a known CVE is present on a scanned or known asset; it is broad and a point in time. Exposure validation goes further by testing whether that vulnerability is actually exploitable, from a realistic attacker starting point, against your current environment, and given current defenses. It turns an assumption into proof, showing the potential impact of exploitation. What is continuous threat exposure management (CTEM)? Continuous threat exposure management (CTEM) is a five-stage framework (scoping, discovery, prioritization, validation, and mobilization) that treats exposure management as an ongoing cycle instead of a periodic audit. It’s designed to limit the impact of long-lived or reintroduced exposures, such as decade-old CVEs. How does CISA’s BOD 26-04 change vulnerability remediation requirements? CISA’s Binding Operational Directive 26-04 moves federal remediation away from flat, fixed-day patch timelines toward risk-based deadlines. KEV status, internet exposure, and the level of attacker control all factor in, with deadlines ranging from 3 to 60 days. For federal agencies, an unpatched KEV entry is now a compliance deadline as well as a security gap. How does SafeBreach help validate exposure to known exploited vulnerabilities? SafeBreach’s CTEM Platform uses adversarial exposure validation to continuously check whether known exploited CVEs are actually reachable and exploitable in your specific environment. Security teams get validated proof instead of an unconfirmed patch status. You Might Also Be Interested In Blog Have Frontier AI Models like Claude Mythos and GPT-5.5 Really Changed the Zero-Day Lifecycle? Read More Solution Brief SafeBreach CTEM Platform Read More Podcast Videos Podcast: BAS, APTs, AEV, and CTEM Explained: A Practical Guide to Cybersecurity Acronyms Listen Now Get the latest research and news

cve漏洞管理报告 @safebreach-blog 原文 ↗

推理基建的缓存隔离假设被打破,多租户 LLM 服务需复核同类设计。

解析大模型推理为何将吃算力的预填充与吃显存带宽的解码拆到不同配置池,以及非对称量化、显存免清零复用与单边 RDMA 这三项提速捷径如何在异常分支下咬合成 Cloudflare GLM-5.3 的跨租户缓存泄露。

llm云缓存泄露 @yage-share 原文 ↗
更多3

仅为提问无实质内容,可观察游戏反作弊与覆盖层挂钩的攻防话题走向。

Hey seen a lot of discussion on hooking trusted overlays but they also seem to be pretty outdated… can anyone point me towards any threads that I should look into? Note: I am not looking to be fully undetected so keep that in mind before commenting something mean

逆向游戏反作弊 @unknowncheats-ac 原文 ↗

与安全无直接关系,仅在评估方法层面可参考其任务设计。

OpenAI November 2, 2026 Research Publication LifeVal Measuring AI progress on the tasks that matter in everyday life. Loading… Share OpenAI’s mission is to ensure that artificial general intelligence benefits all of humanity. For the more than one billion people who currently use ChatGPT every week, we want to build AI that can help with the practical challenges of everyday life, like finding childcare, accessing government services, creating a budget, preparing for school, modifying recipes, ordering groceries, designing workout plans, and more. When someone is looking for help making their life easier, they need more than advice or a chat response. To be a truly helpful personal assistant, ChatGPT should be able to take actions that actually get things done. To measure model performance over time on everyday tasks, we built an evaluation of tasks covering areas where users wanted the most help from AI and personal agents. We used this evaluation to track and drive improvement in model performance on everyday tasks over the past year. Background To identify the tasks people value most, we started with two large studies that track how people spend time in both the U.S. and internationally: the American Time Use Survey ⁠ (opens in a new window) and the Multinational Time Use Study ⁠ (opens in a new window) . Together, these datasets provide a detailed catalog of everyday life tasks that people do across different demographics, times of day, and seasons. To understand where people would most value AI assistance, we expanded on these existing surveys with a new demographically representative survey of 3,995 respondents. For each task, we asked a variety of questions to identify the areas where everyday people most wanted help from a personal assistant. The resulting tasks span the most important areas that people want help with, including caregiving, health, education, household administration, personal finances, shopping, and accessing government services. Individual task examples Below are some examples of how our current models perform on practical workflows from everyday life. Enroll in the right kindergarten after moving 93.1% Compare kindergartens near our new home and shortlist the best fits. I’ve narrowed it to three kindergartens and organized the enrollment details. Two preferences still need your call. Want to review the shortlist? Manage volunteers ahead of a flood 98.0% Help me organize our community’s flood preparation volunteer force for four neighbors needing floodproofing assistance. I’ve prepared a simple action plan to help volunteers assess the sites mentioned and help them prepare for upcoming severe weather. Find a new home for my family 100% Find a home in Luzerne County that fits my budget, has a separate living unit, and keeps my commute short after school drop-off. Add the deed to my cart so I can review it. I found one home that meets your requirements. I’ve calculated the transfer tax and added the deed to your country records cart to review. Watch a game with grandma from far away 96.3% Set up an easy way for grandma and me to watch the game together remotely. The easiest watch setup is ready, with a backup if the stream fails. Want me to walk you through it with grandma? Meal prep for a parent adjusting to a new diabetes diagnosis 90.4% Help my parent adjust to a new diabetes diagnosis with a practical care and meal plan. The first-week plan is ready, including meals, reminders, and questions for the care team. Want to review it together before you share it with your parent? Our latest models score very highly on many LifeVal tasks. See more Over the past year, we’ve trained our models to become much faster, more capable, and more aligned on browser use, computer use, and connectors tasks. Enabling these capabilities in ChatGPT also expands the types of tasks the model can help with, helping move ChatGPT beyond a chat tool and towards a truly agentic personal assistant. Why this matters To ensure AGI benefits all of humanity, we need to measure and improve model progress on what matters in the lives of everyday people. That means going beyond academic assessments of model knowledge and actually measuring how well AI models can help people navigate decisions, manage everyday responsibilities, and get meaningful things done. We are hopeful that LifeVal represents a step towards that goal. 2026 Author OpenAI Keep reading View all How we detect frontier model misalignment Security Oct 29, 2026 Working with academic communities to build scientific understanding Research Oct 26, 2026 ChatGPT’s expanding world of work Research Oct 26, 2026

llm基准 @openai-research 原文 ↗

偏商业与架构分析,仅内网延迟与快照能力对部署选型有边际参考。

Supabase 宣布开启 Compute 私有内测,在数据库同机房内网提供容器托管。本文剖析免后端客群撞上 400 秒超时后的商业防御逻辑与 Deno 并入 Cloudflare 引发的底座接替压力,并结合 Dreambase 实测区分哪些场景是防流失补丁、哪些场景能真正吃到 2.9 毫秒内网与秒级内存快照的红利。

云baas @yage-share 原文 ↗