工业控制漏洞研究
NetVar13372026-10-26查看 361复制 1824
在最新稳定、广泛部署的 SCADA/IIoT、PLC、HMI、historian、工程工作站、OPC、工业网关、远程访问设备、安全系统或设备管理产品中寻找新漏洞时使用。优先关注管理、协议、固件更新、项目文件和控制器工作流中的服务端 RCE 和信任边界缺陷,使用安全模拟和可复现的 ...
📘如何使用这个技能
- 1复制技能内容·点击上方「Copy SKILL.md」复制全部内容,或 Download 下载到本地
- 2打开你的 AI 客户端·支持 Claude Code、Trae、Cursor、ChatGPT、DeepSeek、Kimi 等任意 AI 助手
- 3粘贴为系统提示/知识文件·Claude Code:放入 .claude/skills/ 目录;普通对话:作为首条消息发送并说「请严格按这份技能文档执行」
- 4直接提出你的需求·AI 会按技能定义的工作流、格式和约束来处理,无需再解释背景
💡 提示:技能生效后,每次对话只需一句需求,AI 会按 SKILL.md 的流程执行。
SKILL.md8757 chars
--- name: industrial-control-vulnerability-research description: "Use when hunting new vulnerabilities in latest stable, widely deployed SCADA/IIoT, PLC, HMI, historian, engineering workstation, OPC, industrial gateway, remote-access appliance, safety system, or device-management product. Prioritizes server-side RCE and trust-boundary flaws in management, protocol, firmware-update, project-file, and controller workflows using safe simulation and reproducible ..." version: 1.0.0 license: MIT metadata: package: unleash-skills author: NetVar1337/unleash category: security upstream: C:\Users\Admin\.agents\skills\industrial-control-vulnerability-research\SKILL.md --- > Bundled with Unleash skills pack. Source: C:\Users\Admin\.agents\skills\industrial-control-vulnerability-research\SKILL.md # Industrial Control Vulnerability Research ## Target Priorities Prefer broadly deployed products with one or more: - internet/enterprise-facing remote access, VPN, web, OPC, MQTT, or management service; - engineering workstation that ingests untrusted project/device data; - HMI/historian/SCADA server running with high privilege; - gateway translating untrusted IT protocols into control protocols; - centralized fleet/device management and signed update distribution; - default credentials are not the sole issue; - reproducible latest firmware/software available for lab validation. Focus first on RCE and enterprise/server compromise. Treat process manipulation or safety impact as separate downstream consequences after software-boundary proof. ## Phase 1: Exact Product and Deployment Identity Record: - vendor, model, hardware revision, region/SKU; - latest stable firmware and release date; - bootloader, OS/kernel, web stack, protocol libraries, and application versions; - engineering/HMI/server software and patches; - licensed modules and default-enabled services; - field deployment evidence and typical topology; - update path and whether firmware differs by hardware revision; - hashes and provenance for firmware/installers. Do not assume the highest version string applies to every hardware revision. Verify the vendor's current image for the exact model. Completion criterion: the lab target matches a currently deployed hardware/software line and latest stable release. ## Phase 2: Build a Safe Digital Twin Use, in order: 1. vendor simulator/emulator; 2. isolated hardware bench with no process connection; 3. extracted service in a container/chroot/QEMU system emulation; 4. protocol mock implementing only required state; 5. full controller/HMI pair on an isolated switch. Instrument power/reset recovery and keep a known-good firmware restore path. Stub physical outputs or use simulation mode. Capture baseline traffic before mutation. Completion criterion: crashes and state changes cannot affect production or physical processes, and recovery is tested. ## Phase 3: Attack-Surface Map ### Enterprise/management plane - web/API, VPN, SSH, proprietary admin services; - OPC UA/DA gateways, MQTT brokers, database and historian interfaces; - update, backup, restore, project import/export; - fleet enrollment, certificates, licensing, and remote support; - report generation, alarm/event ingestion, and connectors. ### Control/field plane - Modbus/TCP, DNP3, IEC 60870-5-104, IEC 61850, EtherNet/IP/CIP, PROFINET, BACnet, vendor protocols; - discovery, routing, diagnostics, firmware download, program upload/download; - cyclic/acyclic messages and fragmented transfers; - controller-to-engineering-workstation callbacks. ### File and supply interfaces - project files, ladder/function-block assets, symbol databases; - recipes, trends, reports, HMI graphics, scripts, fonts/images; - firmware packages, modules, drivers, and device descriptions; - removable media and automatic project synchronization. For each surface record default exposure, auth, parser, process privilege, safety role, and updateability. ## Phase 4: Protocol-State Research Capture valid sessions and model: ```text discovery -> capability negotiation -> session/auth -> object addressing -> read/write/execute/download -> completion/error -> disconnect/recovery ``` Mutate with state awareness: - length/count/fragment relationships; - object class/instance/attribute identifiers; - transaction/session sequence numbers; - duplicate, reordered, late, and unsolicited messages; - connection teardown during async work; - route/path nesting and encapsulated protocols; - compression and certificate structures; - diagnostic and maintenance opcodes; - version/capability downgrade. Use grammar-aware clients and snapshots. Separate read-only parser campaigns from state-changing commands. Completion criterion: every mutation is replayable and labels whether it can alter controller state. ## Phase 5: Engineering-Workstation and Server RCE Trace untrusted data arriving from: - controller/device discovery responses; - uploaded project/symbol/configuration; - historian/alarm records; - imported project archives; - web dashboards and report templates; - device-description/plugin packages; - remote support bundles. High-yield sinks: - native parsers and image/font/document libraries; - unsafe deserialization; - script/template/expression engines; - archive extraction and path handling; - dynamic plugin/library loading; - database queries and command helpers; - privileged service IPC from unprivileged UI. A malicious field device that compromises the engineering workstation is a valuable server-side/client-of-device boundary even without internet exposure. ## Phase 6: Firmware and Update Trust Inspect: - signature coverage and verification order; - encrypted but unauthenticated packages; - rollback/downgrade controls; - signed outer archive with unchecked inner files; - hardware-revision and model binding; - bootloader recovery/update channels; - update extraction paths, symlinks, permissions, and hooks; - management server redistribution to fleets; - delta update integer/parsing paths. Distinguish signature bypass from parser RCE before verification. Both can be high impact but require different evidence. ## Phase 7: Embedded Service Analysis For extracted firmware: - inventory architecture, filesystem, init/services, listeners, users, capabilities, and writable roots; - identify third-party versions and vendor patches; - emulate individual daemons with realistic config and filesystem; - fuzz parsers in-process when semantics are preserved; - test on hardware after minimization; - inspect watchdog/restart and fail-open behavior. Do not claim hardware applicability solely from emulation; verify the shipped binary and configuration on the exact latest image. ## Phase 8: Impact and Resilience Classify separately: - management/HMI/historian code execution; - controller code execution; - credential/certificate compromise; - cross-zone pivot; - fleet update compromise; - loss of view/control; - persistent device denial; - process/safety consequence. For the PoC, stop at a harmless marker, controlled crash with automatic recovery, or simulated output. Record watchdog and reboot behavior. ## Variant Hunting Search: - product families sharing protocol stacks; - engineering software and device firmware copies of the same parser; - old hardware branches missing newer fixes; - gateway translations that validate only one side; - project import and controller upload implementations; - management-server and local-tool versions of update logic; - OEM-rebranded devices and licensed third-party components; - recovery bootloader and normal firmware parsers. ## Common Pitfalls 1. Testing latest software against old device firmware. 2. Inferring deployment from internet scans alone. 3. Mutating live control networks or physical outputs. 4. Calling default credentials a novel software vulnerability. 5. Ignoring malicious-device-to-engineering-workstation paths. 6. Assuming encrypted firmware is authenticated. 7. Claiming hardware impact from an emulated daemon only. 8. Mixing software compromise with speculative safety impact. ## Verification Checklist - [ ] Exact model/hardware revision/latest firmware proven - [ ] Deployment evidence and typical topology documented - [ ] Safe digital twin and restore path verified - [ ] Management, field, and file surfaces mapped - [ ] Stateful protocol harness reproduces valid sessions - [ ] State-changing tests separated and controlled - [ ] Engineering/server hostile-input paths traced - [ ] Update signature and model-binding boundaries tested - [ ] Finding reproduced on shipped latest image/hardware where required - [ ] Impact, physical consequence, and resilience reported separately - [ ] Novelty gate passes