LocalLanView holds a map of your whole network: every device, what it runs, and when it's around. Treat its database and console accordingly.
| Threat | Mitigation |
|---|---|
| Someone else on the LAN opens the console | Binds to 127.0.0.1 by default. Binding anywhere else, with -addr or the LAN access setting, requires a password you set yourself (12+ chars, checked for strength). LAN access always uses HTTPS; -addr prints a warning unless TLS is on. |
| Weak or default credentials | No default password. First run generates a random one, prints it once, and stores only an argon2id hash. |
Password leaking via ps or shell history |
No -password flag. Use -password-file, LOCALLANVIEW_PASSWORD or the prompt. |
| Brute-force login | Per-client lockout after 10 failures in 15 minutes (IPv6 clients by /64), a global lockout after 50 that doesn't apply to logins from the machine itself, plus growing delays. Attempts count as soon as they start, and only two password hashes run at once. |
| Cross-site request forgery | SameSite=Strict cookies, a per-session CSRF token on every state-changing request, and an Origin check. Login only accepts JSON. |
| DNS rebinding against the console | On loopback, requests with a non-loopback Host header are rejected. On the LAN, the Host must be an IP address, this machine's name, a TLS certificate name or one passed with -allow-host. |
| XSS through device names, banners, TXT records | The UI never uses innerHTML for data. Strict CSP: script-src 'self', no inline script or style. |
| Formula injection in CSV export | Cells starting with = + - @ are prefixed with '. |
| Other local users reading the database | Data dir is 0700, files 0600, process umask 077 (Unix). On Windows it lives in the user's profile. |
| Running as root longer than needed | Raw sockets are opened first, then the process drops back to the user who ran sudo, doas or pkexec, or to the -run-as account. Started as root with neither, it warns that it's staying root. On Linux, setcap cap_net_raw+ep avoids root entirely. |
| Data leaving the network | No telemetry, background update checks or CDN assets. GitHub is contacted only when you press Check for updates in Settings. HTTP clients that talk to the LAN ignore proxy settings. SSDP descriptions are only fetched from the device that announced them. Webhooks must point at loopback/private IPs. |
| A tampered update | Updates are only installed when you ask. The binary is fetched over HTTPS from the latest GitHub release, must match that release's SHA256SUMS, and must report the expected version before it replaces the running one. Both files come from the same release, so this guards against corrupt downloads, not a compromised GitHub account. |
| Probing things you didn't mean to | All probes are rate limited and restricted to the attached subnet (largest /16). -subnet can only narrow it. |
- Plain HTTP on the LAN. If you bind to a non-loopback address without
TLS, the password and data cross the network unencrypted. Use
-tls-self-signedor your own certificate. - A compromised machine. Anyone who can run code as your user can read the database.
- Active attackers on the LAN. Devices can lie in mDNS, SSDP and banners. Treat names and types as hints, not proof. The "MAC changed" alert can catch some ARP spoofing but isn't an IDS.
- Sessions live in memory and are lost on restart (you log in again).
Only scan networks you own or are authorized to administer. Port scanning someone else's network can break their policies or the law.