⚡Megatron LLM Hub
全部
Website

rss-website

头条1

容器基础镜像常在默认配置下受影响,云环境可被未认证攻击者直接拿到 root。

Introduction The cybersecurity landscape recently witnessed a highly critical vulnerability dubbed “SSHole” (CVE-2026-123158). This flaw represents a severe authentication bypass issue that primarily impacts lightweight, container-friendly operating systems like Alpine Linux. By exploiting an unintended collision between several independent security mechanisms, unauthenticated remote attackers can bypass SSH authentication entirely. In specific cloud environments, this seamlessly translates to immediate, unauthorized root access, making it one of the most critical threats to modern cloud infrastructure and containerized workloads. Because Alpine is the foundational base image for millions of Docker containers and Kubernetes deployments worldwide, the potential attack surface is staggering. Furthermore, the exploit leaves minimal traces in standard access logs, making post-incident forensic analysis incredibly challenging for security teams trying to identify unauthorized intrusions before massive data exfiltration occurs. Learning Objectives By reading this vulnerability analysis, you will be able to: Grasp the foundational security concepts that inadvertently collided to create the SSHole vulnerability. Understand the specific technical interaction between musl libc and OpenSSH. Identify the exact system configurations and conditions that make an environment vulnerable. Apply immediate mitigation techniques and permanent remediation steps to secure affected systems. What is SSHole? A Critical Root Vulnerability in Alpine Linux Systems (CVE-2026-123158) SSHole (CVE-2026-123158) is a critical authentication bypass vulnerability that emerges not from a single coding error, but from the dangerous overlap of three distinct security features. In Linux systems, user accounts that are disabled or do not have a password (meaning they should only be accessed via SSH keys) are securely marked with an asterisk ( * ) in the password field of the /etc/shadow file. Under normal circumstances, this mechanism guarantees that no text-based password can ever match the stored hash, theoretically locking the account down against brute-force attacks. However, the vulnerability occurs when OpenSSH’s defense mechanism against user enumeration timing attacks interacts poorly with a Denial of Service (DoS) protection feature built directly into the musl libc library. This obscure edge-case interaction essentially tricks the SSH daemon into believing that an attacker has successfully provided the correct password for a completely locked-down account, bypassing the cryptographic barrier entirely. The implications of this flaw are profoundly amplified in modern cloud architectures. Because administrators heavily rely on default accounts without passwords to streamline automated deployments and provisioning scripts, the attack surface expands dramatically. Once the authentication layer is bypassed, the system inherently trusts the rogue connection, paving the way for complete infrastructure compromise without a single valid credential. It is classified as an Authentication Bypass vulnerability with a critical CVSS score of 9.8. It specifically targets the flawed interaction between the musl libc cryptographic functions and OpenSSH’s internal daemon logic. It primarily affects system accounts that have been explicitly disabled or locked using the * shadow hash indicator. It requires absolutely no prior knowledge of the target system’s credentials, private SSH keys, or active user behavior. +--------------+ +-------------------------+ | | 1. Send 257-byte password | | | Attacker | ----------------------------> | OpenSSH Daemon | | | | | +--------------+ +-------------------------+ | | 2. Pass input to crypt() V +------------------------------------+ +-------------------------+ | | <------ | | | /etc/shadow File | 4. Match| musl libc | | | '*' | | | Account: alpine | with | 3. Reject oversized input, | Password Hash: '*' | '*' | return '*' as error | +------------------------------------+ +-------------------------+ | | 5. MATCH FOUND! V [ Access Granted -> Root via doas ] Technical Detail: How the Vulnerability Works The mechanics of the SSHole vulnerability can be broken down into a sequence of events triggered by a meticulously crafted, anomalous login attempt. To prevent CPU exhaustion attacks, the crypt() function within the musl libc library is designed to reject passwords that exceed 256 bytes in length. When it encounters such a massive input, it refuses to process the expensive hashing algorithm and instead safely returns a single asterisk ( * ) as an error indicator to the calling application, aiming to drop the load immediately. Concurrently, OpenSSH is designed to prevent attackers from guessing valid usernames based on server response times. If an attacker tries to log into an invalid account or an account without a password, OpenSSH still performs a “dummy” password verification in the background using the password salt of another valid user on the system. This ensures that every login attempt takes the exact same amount of time, masking the existence of specific users and thwarting enumeration techniques. The fatal collision happens when an attacker targets an account that has no password (like the default alpine user, which possesses a * in the /etc/shadow file). The attacker intentionally sends a password longer than 256 bytes. The musl libc library rejects the long password and outputs the error code * . OpenSSH’s faulty validation logic then blindly compares this returned * with the actual hash stored in the shadow file (which is also * ). Because the two values perfectly match, OpenSSH interprets this as a successful cryptographic verification. Trigger: The attacker initiates a standard SSH connection but provides a manipulated password string exceeding the 256-byte limit. Library Response: The musl libc library intercepts the oversized string, aborts the hash calculation, and returns the * error token. Daemon Logic: OpenSSH proceeds with its timing mitigation routine, improperly comparing the returned error token directly against the shadow file entry. Escalation: Access is mistakenly granted to the default account, which subsequently leverages its pre-configured doas rules to instantly achieve root privileges. Affected Software & Plugins The scope of CVE-2026-123158 is heavily restricted by the specific architectural choices of the underlying operating system and its authentication daemons. Unlike widespread vulnerabilities that affect the ubiquitous glibc library, this flaw is strictly confined to systems that rely on the lightweight musl libc implementation. This makes Alpine Linux, Void Linux (the musl variant), and various embedded firmware solutions like OpenWrt the primary targets for this sophisticated attack vector. Furthermore, the vulnerability requires a very precise configuration state within the OpenSSH daemon itself. Systems utilizing Pluggable Authentication Modules (PAM) are naturally immune because PAM overrides and entirely replaces the faulty internal OpenSSH validation logic. The attack only succeeds in environments where basic password authentication is enabled, PAM is disabled, and the specific shadow file conditions are met simultaneously. Operating systems utilizing musl libc instead of standard glibc (e.g., Alpine Linux, OpenWrt, Buildroot). OpenSSH daemon versions prior to 10.6p2 that were compiled or configured without Pluggable Authentication Modules (PAM) support. System configurations explicitly allowing standard password authentication ( PasswordAuthentication yes in the sshd_config file). Systems containing at least one disabled or passwordless account ( * ) alongside at least one valid hashed account to facilitate the timing mitigation phase. Conclusion The SSHole vulnerability (CVE-2026-123158) serves as a stark reminder that isolated security mechanisms can sometimes combine to create catastrophic systemic failures. It demonstrates how a Denial of Service protection rule in one underlying library and a timing attack mitigation feature in a network daemon can unwittingly collaborate to bypass authentication entirely. This “perfect storm” of overlapping logic flaws highlights the extreme complexity and fragility of modern software stacks. System administrators operating Alpine Linux or other musl -based systems must take immediate and decisive action to secure their infrastructure against this threat. To permanently remediate the issue, OpenSSH must be updated to version 10.6p2 or the subsequently patched version provided by the respective OS maintainers. Crucially, updating the software packages is not enough on its own; a full restart of the SSH daemon is mandatory to clear the vulnerable code from the active system memory. If immediate patching is not feasible due to strict production constraints or uptime requirements, administrators can apply a highly effective temporary mitigation by altering their SSH daemon configuration. By explicitly setting PasswordAuthentication no in the sshd_config file and relying strictly on cryptographic keys, the attack vector is completely neutralized. Ultimately, adopting a Defense in Depth strategy remains the best safeguard against such unpredictable, multi-layered vulnerabilities. Simple CTF TryHackMe Walkthrough Improving Password Security with Machine Learning What is AS-REP Roasting Exploitation: A Comprehensive Guide Exploiting Zero-Day Vulnerabilities in SonicWall SMA Prior to Disclosure Linux Kernel Vulnerability CVE-2024-26582: Local Privilege Escalation and Root Shell Analysis https://nvd.nist.gov/vuln/detail/cve-2026-123158 https://www.tenable.com/cve/CVE-2026-123158 https://socprime.com/blog/cve-2026-123158-openssh-password-authentication-bypass-affects-alpine-linux/ The post Vulnerability Analysis: CVE-2026-123158 “SSHole” – Critical Root Vulnerability in Alpine Linux Systems appeared first on DenizHalil - Professional Cybersecurity Consulting and Penetration Testing .

cve认证绕过云容器 @denizhalil-blogs 原文 ↗
推荐3

给出可直接搬进 agent 主循环的工程取舍,避免为多代理架构白付成本。

Surge AI 与 UCLA FireFlash 项目的同期实测拆解表明:派 17 个子代理轮流开会将单题成本推高至 8 倍却无显著提分,真正经受住跑分检验的是调高单模型思考档位、在主循环守住先写失败测试与回合末目标核对两个轻量关卡,并在底座配齐慢命令转后台、长输出落盘预览与追加式摘要保缓存。

ai_agentllm工程实践 @yage-share 原文 ↗

涉及多智能体权限隔离与提示缓存命中的落地账,内网部署选型可直接参照。

月之暗面发布的 Kimi Hosted Agents 与智谱 Z Managed Agents 双双放弃跟进 OpenAI 的单端点续写接口,转而逐行复刻了 Anthropic 的四元控制面协议。本文拆解三层企业内网混合执行方案,并结合官方行业样板间算清多智能体物理权限隔离与 95% 提示词前缀缓存命中率这两笔实在账。

ai_agentllm架构云 @yage-share 原文 ↗

给出驱动层拦截设备枚举查询与遥测上报的思路,可作反检测研究参考。

Hey everyone, Looking into implementing a custom approach for BSSID/SSID spoofing and spoofing HWID at the Ring 0 level. I've been looking at NDIS LWF drivers for intercepting Wi-Fi enumeration queries and handling system telemetry, but I'm trying to figure out the most reliable way to handle modern checks without triggering flags. Has anyone here experimented with Native 802.11 request manipulation or building custom filter drivers for this purpose? Would love to exchange some ideas or discuss how to properly structure this. Feel free to reply here or check my profile.

内核驱动逃逸反作弊 @unknowncheats-ac 原文 ↗
更多1

01 Reasoning through geometric problems A little over a year ago, we began testing how well Claude could design basic electronic circuits. We collaborated with Diode Computers , 1 a company that designs and manufactures circuit boards, to train Claude on simple tasks—for example, on reading a chip’s datasheet and working out which parts a chip needs in order to run. The models we experimented with back then couldn’t reliably lay out a circuit board; they also struggled to design components that fit together. Now, Claude can both design and prototype electronic hardware from start to finish. Engineering an electronic device requires working through many complex steps and practical constraints. You have to not only design and simulate a circuit board, but also reason about the board’s physical layout and how it will behave, model the mechanical components, and write firmware—all while keeping track of which parts are readily available and what a factory is set up to build. To see how newer Claude models have improved in this area, we collaborated with Diode again, this time asking Claude Fable 5.2 to design three electronic devices: a speakerphone, a voice-assistant hub, and a handheld virtual pet. With minimal prompting, Claude was able to use geometric reasoning to lay out each circuit board, anticipate manufacturing constraints, and make sure its designs would work within them. It revised its own work, modeled its designs in computer-aided design (CAD) software, and, finally, directed human engineers to build prototypes that worked without any fixes to their circuitry. For each design, we began by giving Claude a simple prompt that outlined what kind of device to build and its main hardware requirements. The prompt for the speakerphone, for example, asked for “a puck that plugs into a computer over USB-C and includes microphones and a speaker, supporting Bluetooth with a PCB antenna.” Claude proceeded to write the device’s specifications, draw clean visual schematics, choose where to place the components on a printed circuit board (PCB), route the copper connections between them, design a 3D-printable enclosure, code the firmware, and prepare a parts list. Diode assembled the circuit boards and ran the firmware, under Claude’s direction. The final result was a prototype of a speakerphone that still works after weeks of use. Claude went through a similar process to build the voice-assistant hub and the virtual pet, and both are also still in good working order. Because the devices are early prototypes, we don’t yet know how they’ll hold up with continued use or whether they would pass commercial certifications. The three devices built from Claude’s designs, from left to right: a battery-powered handheld pet, a home voice-assistant hub, and a USB speakerphone. Claude engineered the electronics and the enclosures and wrote firmware for each one. Reasoning through geometric problems The most substantial difference between Fable 5.2 and the models we used in our earlier tests was in how well Fable 5.2 reasoned through complex geometric problems, such as where to place components on a circuit board and how to route the copper connections between them. Earlier models had struggled to decide how to arrange basic components—placing supporting parts too far from the chips they served, for example, or orienting connectors incorrectly. Fable 5.2, by contrast, produced more compact, practical layouts. It routed the most important connections first: those that carry high-speed signals or high currents, or that are sensitive to noise. For example, for the voice-assistant hub, Claude started by routing the USB and Ethernet connections, which transmit signals along pairs of tracks—thin strips of copper on the board. It routed one pair of tracks at a time, keeping the two tracks in each pair the same length and correctly spaced. It also added tuning patterns—small zigzags that lengthen a track so that signals arrive at the same time—and arranged the parts of the power supply to make it more efficient. To demonstrate how much better Fable 5.2 is than earlier models at laying out circuit boards, we asked Diode to run PCB Gym—an internal benchmark that scores models based on how well they can place components and reroute connections on board designs whose layouts have been removed. Fable 5.2 scored 69% on the benchmark, compared with 25% for Fable 5 (Figure 2). It also performed better on BenchCAD 2.1 , a public benchmark that tests whether models can rebuild 3D parts, assemblies, and circuit boards from engineering drawings and rendered images. Fable 5.2 scored 83.5%, nearly double the score of its predecessor, Claude Fable 5.1. Scores plotted against total API cost for recent frontier models on Diode PCB Gym, a benchmark that instructs models to place components and route connections on circuit boards whose layouts have been removed. Revising designs Earlier models rarely revised their own work when designing PCB layouts, even when it became clear that their placement choices would never let them finish routing the connections. Over the past year, Diode designed many boards with these models that were 95% complete but still unusable. When Claude noticed one of its designs wasn’t working, it would change course without being prompted to do so. For instance, when designing the speakerphone’s circuit board, which has four layers of copper, it reassigned signals to different pins on two of the chips. This allowed the tracks under the speakerphone’s touch-sensitive controls to stay on a single layer without crossing one another. Later, when the back of the board became too crowded, Claude swapped the signals on interchangeable pins and laid out a whole corner of the board again, rather than patching the problem with extra jumps between layers. And when placing tracks in its design file for the voice-assistant hub, Claude noticed that the copper connections were “getting too tangled to trace visually” and reasoned, “I should just rip these traces [tracks] and start fresh.” It proceeded to delete and reroute copper connections dozens of times until it found an arrangement that worked. The routing for the voice-assistant hub before and after Claude reorganized the connections. Left: the microphone’s power track (blue lines) loops around the ground ring at the sound port, a front-side track (red) crosses the ring, and a via—a small hole that connects layers of copper—is placed between the microphone’s pads. Right: the same area after Claude rerouted the power track, with the ring clear and grounded. Claude’s reasoning for redoing the version on the left: “That’s risky placement under a bottom-port MEMS mic, even if clearance is technically fine, so I want to redo the local routing cleanly for both identical mic footprints.” Using software and documentation When a design decision for one of the devices depended on the details of a particular part, Claude often checked that part’s datasheet—a technical document published by the manufacturer. When asked whether the handheld pet would work without a battery, Claude read the charger’s datasheet and answered that it would run on USB power but predicted that the charging light would flicker, because the charger keeps recharging a small capacitor in the battery’s place. 2 To route the connections on the voice-assistant hub, Claude started with the design software Diode provided, which let it run the router (a tool for drawing tracks) built into KiCad, a widely used circuit board design program. KiCad’s router drew one track at a time between two points Claude chose. But the hub had long tracks running from a 100-pin connector, and Claude judged that routing these one by one would take too much step-by-step guidance. Instead, Claude wrote its own router: an algorithm that divides the board into a grid and searches it to find a clear path for each track. Within about half an hour, Claude had a router that could lay each long track in roughly a second. When Claude spotted an audio clock track running too close to the Ethernet tracks, it adjusted the router to leave more room around sensitive tracks. Keeping track of constraints and downstream effects Decisions made in one part of a design have downstream effects, which Claude kept track of. While designing each device’s enclosure, it wrote code to check that a 3D model of the circuit board would fit inside the enclosure without any components colliding. For example, it found that the speakerphone’s board couldn’t be lowered straight into the enclosure, because the mute switch would hit the back wall. Claude worked out that the board would fit if it was tilted by about 25 degrees during installation. It tested this in its 3D model and added the step to the assembly instructions. When designing the enclosure for the voice-assistant hub, Claude calculated that a 25 mm-long microphone port—the opening that lets sound reach the microphone—would resonate near 3.4 kHz. That is “right in the voice band,” it reasoned, “which is not ideal though still usable for voice assistant purposes. A better approach is a wedge-shaped enclosure that’s tall at the rear and steps down low at the front.” The wedge shape shortened the ports to about 10 mm, which raised the resonant frequency to about 8.9 kHz, at the very edge of the range used for speech. Claude also wrote firmware for the speakerphone that dims its ring of LED lights as the audio becomes louder, to keep the device under the electrical current limit Claude had designed into the board. Claude’s designs for the home voice-assistant hub and the finished prototype. Claude wrote the code that modeled the case in CAD, laid out the circuit board, and wrote the firmware. Designing for real-world manufacturing In electronics design, engineers have to work within the constraints of the real world. That is, they have to make sure a factory can actually make the thing they want to build. They do this by running two kinds of checks: design-for-manufacturing (DFM) checks, which compare a circuit board’s geometry with the factory’s manufacturing limits, and sourcing checks, which confirm that their distributors have the parts in stock. Claude ran these checks itself and sorted the results into problems that needed fixing and flags it could ignore. Its checks for the voice-assistant hub flagged six slivers of solder mask (the protective coating on a circuit board) as too thin to manufacture. Claude fixed them by widening the openings in the solder mask, without moving any parts. The checks also flagged a connection between two grounds (shared return paths for electrical current) as an error, because the automated checker hadn’t recognized that the connection was deliberate. Claude correctly reasoned that it wouldn’t affect performance and ignored the flag. Diode manufactured all three boards directly from Claude’s design files, without any back-and-forth over adjustments. Factories and engineers often have to work through multiple rounds of manufacturing changes, and the Diode team told us they were surprised that they didn’t have to do this with Claude. Layer 1: front Layer 2: inner 1 Layer 3: inner 2 Layer 4: back The speakerphone’s four copper layers. The dashed circle marks the touch wheel, where copper electrodes sense the user’s finger. Moving a track from one layer to another requires drilling a hole in the board, and under the touch wheel, that hole would go through an electrode. To avoid this, Claude kept each signal track under the wheel on a single layer, and it chose the chips’ pin assignments so that those tracks wouldn’t need to cross one another. It also kept the area around the Bluetooth antenna (the zigzag at the top) clear of other copper and holes on every layer, so nearby metal wouldn’t weaken its signal. Layer 1: front Layer 2: inner 1 Layer 3: inner 2 Layer 4: back The voice-assistant hub’s four copper layers. Claude added two areas of grounded copper to the power layer (green), directly behind most of the tracks on the back of the board that carry high-speed signals. This gives the signals a nearby return path, which helps reduce interference. The speakerphone’s circuit board can’t be lowered straight into the enclosure, because the mute switch’s knob would hit the back wall (red). Claude worked out how to install the board by tilting it about 25 degrees and then swinging it flat, and added this step to the assembly instructions. Limitations and next steps All three of these devices are still prototypes: they haven’t been through electrical validation, antenna tuning, thermal testing, or certification, and they weren’t designed for mass production. Since Claude wasn’t connected to test equipment or to the machines that make and assemble circuit boards, engineers at Diode had to physically build and test the prototypes. Even so, over the past year, Claude has improved substantially at reasoning through decisions about component placement and routing, and it can now design electronics with minimal human guidance and help people build the prototypes. Other models are making progress here, too: OpenAI recently wrote about using its models to design a handheld device. Work like this could help engineers build and test new hardware designs faster. Next, we plan to investigate whether Claude can design denser, faster, and more complex boards, check them in simulation, and learn from measurements taken on the finished devices. Footnotes 01 Anthropic has a commercial relationship with Diode. 02 This was one of the few instances where the Diode team prompted Claude (“Will it work if there is no battery?”). Overall, their prompting was minimal and limited to short directives (e.g., “clean up the schematic” and “continue”). Authors & acknowledgements Acknowledgements This post is a collaboration between researchers and engineers at Anthropic, including Mike Cretella, Connor Jennings, and Brent Yi, and Craig Chen and the team at Diode Computers. Anthropic has worked with Diode electrical engineers on PCB design since 2025. Editorial support from Xander Balwit, Tom Chivers, Rebecca Hiscott, and Caitlin Love.

llm推理硬件 @anthropic-research 原文 ↗