Skip to content

Security: vicanso/zedis

Security

SECURITY.md

Security Policy / 安全策略

🇬🇧 English

Supported versions

Security fixes are applied to the latest release line only.

Version Supported
0.17.x (latest) ✅
< 0.17 ❌

Reporting a vulnerability

Please do not report security vulnerabilities through public GitHub issues.

Instead, report privately via GitHub's Report a vulnerability (the repo's Security → Advisories tab). Please include:

  • the affected version and platform,
  • a description of the issue and its impact,
  • steps to reproduce (and a proof of concept if possible).

We aim to acknowledge your report within a few days and will keep you updated as we work on a fix, coordinating disclosure timing with you.

Scope

Zedis is a local desktop client — it stores connection secrets encrypted at rest, keeps metadata (tags, notes, favorites, history) in a local file, and its only unprompted outbound network call is the startup update check against GitHub Releases, falling back to the Gitee mirror when GitHub is unreachable (off with one Settings switch); the optional AI analysis talks only to the endpoint you configure. The full privacy statement is in the README's Code signing policy. Reports are especially welcome around:

  • connection-secret handling and the at-rest encryption,
  • TLS/SSL and SSH-tunnel handling,
  • the custom script viewer (which runs local shell commands),
  • the optional AI analysis (what leaves your machine, and where).

The self-hosted web version adds a server of its own, zedis-bridge: it holds the server list and its secrets, signs accounts in and forwards their commands, so it is reachable over the network by design. Reports about it are just as welcome, in particular:

  • sign-in, saved logins and the trusted-proxy header (--trusted-header / --trusted-proxy),
  • a read-only account, or a :ro entry, getting a write through — the MCP entry point (/v1/mcp) included,
  • one account seeing another's private server entry, or any secret reaching the browser,
  • the audit log leaving out something it says it records.

Threat model

A Redis server you connect to is treated as untrusted input. Key names and values are chosen by whoever can write to that server, so Zedis never lets them become executable syntax: the custom script viewer passes {KEY} and {VALUE} to the shell through environment variables that expand after the command line is parsed, and every script run is bounded by a timeout and an output cap. A report showing that data from a server can influence what runs on the client machine is exactly the kind we want.


🇨🇳 中文

受支持的版本

安全修复仅应用于最新的发布线。

版本 是否支持
0.17.x(最新) ✅
< 0.17 ❌

报告漏洞

请勿通过公开的 GitHub issue 报告安全漏洞。

请通过 GitHub 的 Report a vulnerability(仓库 Security → Advisories 标签)私下报告,并尽量包含:

  • 受影响的版本与平台;
  • 问题描述及其影响;
  • 复现步骤(如可能,附概念验证)。

我们会争取在数日内确认你的报告,并在修复过程中持续同步进展,与你协调披露时间。

范围

Zedis 是本地桌面客户端 —— 连接密钥加密存储,元数据(标签、备注、收藏、历史)只存本地文件,唯一会主动发起的外发网络请求是启动时对 GitHub Releases 的更新检查,GitHub 不可达时回退到 Gitee 镜像(设置里一键关闭);可选的 AI 分析只连接你自己配置的端点。完整的隐私声明见 README 的代码签名策略。以下方面的报告尤其欢迎:

  • 连接密钥的处理与静态加密;
  • TLS/SSL 与 SSH 隧道的处理;
  • 自定义脚本查看器(会执行本地 Shell 命令);
  • 可选的 AI 分析(哪些数据离开本机、发往何处)。

自托管的 Web 版多了一个自己的服务端 zedis-bridge:它保存服务器列表及其密钥、负责账号登录并代为转发命令,因此按设计就是网络可达的。关于它的报告同样欢迎,尤其是:

  • 登录、已保存的登录会话与受信代理请求头(--trusted-header / --trusted-proxy);
  • 只读账号或标了 :ro 的条目让写入通过了 —— 包括 MCP 入口(/v1/mcp);
  • 一个账号看到了另一个账号的私有服务器条目,或任何密钥到达了浏览器;
  • 审计日志漏记了它声称会记录的事件。

威胁模型

所连接的 Redis 服务端被视为不可信输入。键名与值由任何对该服务端有写权限的人决定, 因此 Zedis 不会让它们变成可执行的语法:自定义脚本查看器通过环境变量把 {KEY} 与 {VALUE} 传给 Shell,变量在命令行解析完成之后才展开;每次脚本执行都有超时与输出上限。 如果你能证明来自服务端的数据可以影响客户端上执行的内容,正是我们最希望收到的报告。

There aren't any published security advisories