Skip to content

Canvus security overview

MultiTaction Canvus is software for interactive digital collaboration. It is a self-hosted, on-premise solution: you run Canvus Server inside your own IT infrastructure.

The Data protection facts section at the end of this page answers the questions that security questionnaires usually ask.

Security architecture

Canvus uses a client-server architecture with two main software components: the Canvus client and Canvus Server.

The Canvus client is a native desktop application written in C++ for Windows and Ubuntu Linux. It runs on personal computers, such as desktops and laptops, and on dedicated computers connected to an interactive video wall.

Canvus Server synchronizes the Canvus clients. It stores user accounts, permissions and canvas data in a PostgreSQL database, and uploaded files in its asset folder. The server also serves the web client and the administration pages to browsers.

Data flow

Users connect to Canvus Server with the Canvus desktop client, a web browser or the REST API.

flowchart LR
    C["Canvus client<br/>(desktop)"] --> S["Canvus Server"]
    W["Web browser"] --> S
    R["REST API"] --> S

Canvus Server also opens a small number of outbound connections of its own, for example to check its license, send email or upload a crash report. See Outbound connections.

Data transmission

Browsers, desktop clients and REST API callers all connect to the same server port.

When you use a web browser, the browser and Canvus Server form a single HTTPS connection. TLS encrypts all data, and the server's TLS certificate identifies the server to the browser.

The Canvus desktop client uses its own protocol (not HTTP) inside a single TLS connection. It uses the same certificate and key as the HTTPS connection.

Both kinds of traffic arrive on port 443 by default, and Canvus Server separates them internally. TLS 1.2 is the minimum protocol version the server accepts. While TLS is enabled (ssl-enabled=true, the default), the server refuses unencrypted desktop-client connections and redirects plain HTTP requests to HTTPS.

All three deployment types also enable a redirect from HTTP on port 80 to HTTPS by default (http-redirect-enabled=true). On a Windows standalone install, the installer opens the Windows Firewall for the HTTPS port only, so other hosts reach the port 80 redirect only if you open that port yourself.

The ports used by Canvus Server are:

Port Protocol Purpose
443 TCP HTTPS (external; clients and browsers connect here).
80 TCP HTTP redirect to HTTPS (external; redirects all plain-HTTP requests to port 443).
4443 TCP WebRTC signaling WebSocket between canvus-combined and canvus-media (internal only; not published to the host). WebRTC video streaming is off by default. See WebRTC video streaming.
40000-40100 UDP WebRTC media (external; used only when WebRTC video streaming is turned on).

If an administrator sets ssl-enabled=false, all traffic is unencrypted. Do not do this on a network you do not fully control.

Certificates

TLS needs an X.509 certificate. By default, Canvus Server creates a self-signed certificate when it first starts, so traffic is encrypted without any manual configuration. Browsers and clients cannot verify a self-signed certificate, so users see a warning.

You can install your own certificate and key in PEM format, issued by any Certificate Authority you choose. Do this for every production deployment.

Customer system integrations

  • Email. Canvus Server sends email through your SMTP server. Email is needed, for example, to let users reset a lost password.
  • Single sign-on (OpenID Connect) - Canvus Server can integrate with Microsoft Entra ID, or with another OpenID Connect provider, for user authentication. See Set up single sign-on with Microsoft Entra ID.

Audit trail

Canvus Server records security-related events in an audit trail. Administrators view it in the web UI under Admin > Audit log, and can export it as a .csv file.

The audit trail is also available through the REST API for third-party integrations.

Access control

Canvus provides access control with user accounts and predefined roles. There are two types of user account:

  • Regular users are subject to normal access control restrictions.
  • Administrator users can access all content on the server and manage all features of the installation.

A set of predefined roles restricts access to content. You can also put user accounts into user groups to manage access more easily.

A built-in Guest user account is enabled by default for scenarios where people need easy access to some content without a personal account. While the Guest account is enabled, users can also share canvases through links that need no authentication.

The Guest account can open content on Canvus Server with the Canvus desktop client. In a web browser, the Guest account can contribute content to canvases that have been shared with Anyone with the link can edit access.

You can block the Guest account to disable all anonymous access and make every user sign in.

Authentication

Canvus Server provides two authentication methods:

  • Local password authentication uses passwords stored in the PostgreSQL database to authenticate users.
  • Single sign-on uses OpenID Connect with Microsoft Entra ID or another OpenID Connect provider to authenticate users. SAML was removed in Canvus Server 26.11.1.

By default, local password authentication is enabled. It can be disabled if all users are to authenticate with single sign-on.

Single sign-on uses the OpenID Connect authorization code flow with PKCE (S256). The server creates a random state and nonce for each sign-in, keeps them in short-lived cookies (5 minutes, HttpOnly, SameSite=Lax, Secure over TLS), and verifies the signature, issuer, audience, expiry and nonce of the ID token against the provider's published keys. For Microsoft Entra ID it also checks that the token comes from the configured tenant. The server proves its identity to the provider with a client secret or with a certificate (private_key_jwt). The client secret and the private key of the certificate are stored encrypted with AES-256-GCM in sso-config.json. The 32-byte encryption key is in a separate file, created with owner-only permissions, in a different folder. The admin panel never shows a stored secret or private key again. Administrator rights are never granted by the identity provider.

QR code authentication

QR code authentication keeps users from typing their credentials on a video wall where other people in the room can see them.

With QR code authentication, users scan a QR code on the video wall with their mobile device and perform the authentication on their device. The actual authentication relies on either local password authentication or single sign-on. After authentication on the device, a secret token is used to authenticate the user on the video wall.

QR code sign-in is offered on the sign-in page by default.

Session management

Canvus stores some information on the user's device when they sign in.

  • In the Canvus desktop client:
    • If the user selects Remember me when signing in to a Canvus Server, and the client is not in multi-user mode, the client stores a sign-in token and the user name in the operating system's secret store. Where the operating system has no secret store available, the client keeps them in a file in the user's profile instead. The user can then sign in again without re-entering the user name or password.
    • If persistent browser sessions are enabled (the default), the desktop client stores the user's browser session (cookies, history and so on) on disk inside Canvus. Users who are signed in to Canvus Server (not as Guest) stay signed in to web pages after they close the canvas or restart the client. Each user has their own browser session, which is not shared with other browsers on the same device.
  • In a web browser:
    • If the user selects Remember me when signing in, the browser keeps a sign-in cookie so the user can return without re-entering the user name or password.

The user's password is never stored on the device running the Canvus client.

The browser sign-in endpoints have cross-site request forgery (CSRF) protection.

User credentials

With local password authentication, user credentials are stored in the PostgreSQL database. Passwords are stored as hashes, never in plain text.

When using only single sign-on, no user passwords are stored in Canvus.

Forgotten passwords

With local password authentication, users can reset a forgotten password themselves. The server emails the user a temporary link to set a new password. This needs email to be set up.

Administrators can also reset users' passwords or set them directly.

End-user identification

User sign-up can be disabled for both local password authentication and single sign-on. If disabled, all user accounts must be created by an administrator.

With sign-up enabled, you can require an administrator to approve each new account, and restrict sign-up to email addresses in the domains on an allow list.

With single sign-on, the first sign-in links the user to the existing Canvus account with the same email address, including administrator accounts. When there is no such account and Allow new accounts is on (the default), the server creates one; the domain allow list and administrator approval still apply. Control who can sign in at your identity provider, for example by requiring user assignment on the Microsoft Entra enterprise application.

Data protection facts

This section summarizes, in questionnaire order, how Canvus Server handles data. It describes Canvus Server 26.11.0.

Deployment model

  • Canvus Server is self-hosted. You run it on your own hardware or virtual machines, as a Linux package (Ubuntu 24.04), a Windows installer (Windows Server 2019 or 2022), or containers (Podman).
  • Each deployment is a single server instance with its own PostgreSQL database. There is no clustering or failover. See Health, monitoring and sizing.
  • MultiTaction does not host or operate your Canvus Server.
  • A server without internet access can be licensed with offline activation. See licensing.

Where data lives

All of these are on the server you run:

Data Where
User accounts (name, email address, password hash, admin, blocked and approval flags, created and last sign-in times), groups and permissions PostgreSQL database
Canvases, folders and canvas content PostgreSQL database, with uploaded files in the asset folder
Uploaded files (images, videos, PDFs and other assets) Asset folder ([content] root in mt-canvus-server.ini)
Audit log (time, user, IP address, action, target, details) PostgreSQL database
API access tokens PostgreSQL database, stored as hashes
Daily usage counts with internal user ID numbers (only with a usage-reported license; no names or email addresses) usage.db in the server's data folder
Server configuration, including the SMTP and database passwords mt-canvus-server.ini
License files, TLS certificate and key, session secret Server configuration and data folders
Server and database logs Log folders on the server

See file locations and Server database for the exact paths.

Canvus desktop clients also keep data on the computer they run on: local canvases, which are stored only on that computer, and a cache of the files on shared canvases. To make sure no canvas is stored only on a client computer, such as the PC of a video wall, turn local canvases off with disable-local-canvases=true in the [system] section of the client's mt-canvus.ini. See Turn off local canvases. How long cached files are kept is set by [cache] max-age; see Cache and asset management.

Encryption in transit

  • Browsers, desktop clients and the REST API connect over TLS 1.2 or later on one port (443 by default). TLS is on by default.
  • You choose the certificate: the server's self-signed default or one from your own Certificate Authority.
  • The connection between Canvus Server and its PostgreSQL database is not encrypted, and Canvus Server has no setting to encrypt it. With the bundled database it stays on the same computer (the bundled PostgreSQL listens on the loopback interface only on standalone Linux and Windows installs) or on the private container network. Canvus Server cannot connect to a PostgreSQL server that requires SSL, so external databases that require SSL are not supported. See Server database.
  • SMTP to your mail server can use TLS or STARTTLS, set in the server's email settings.
  • The canvas call travels inside the client connection above. For what that means for privacy, see Canvas call administration.

Encryption at rest

Canvus Server does not encrypt data at rest itself. The database, the uploaded files, the logs and the backups are stored unencrypted by the application. To encrypt them at rest, use full-disk or volume encryption on the server, for example BitLocker on Windows or LUKS on Linux, and on the disk that holds your backups.

On a Linux standalone install, the canvus service user owns the data and the configuration file is mode 600. On Windows, the data folder is limited to SYSTEM and Administrators, and the configuration file to SYSTEM and Administrators with read access for the LocalService account.

Identity and access

  • Local accounts with a configurable password policy (length, character classes, repeats and sequences, password history and maximum age).
  • Single sign-on with OpenID Connect, including Microsoft Entra ID (from Canvus Server 26.11.1, which removed SAML). Multi-factor authentication is provided by your identity provider.
  • Sign-in on a video wall with a QR code and a mobile device.
  • Sign-up controls for local accounts: sign-up on or off, an email-domain allow list, and administrator approval of new accounts. The allow list and approval also apply to accounts created by single sign-on, and your identity provider decides who can sign in.

Retention and deletion

  • Canvus Server has no retention schedule. Canvases, user accounts, audit log entries and daily usage counts stay until an administrator or user deletes them.
  • The server removes some data on its own: an uploaded file once no canvas refers to it any more, expired sign-in tokens and password reset links, and sign-up registrations whose email address was never confirmed.
  • A deleted user account stays in the database as an archived record, so that records which refer to it remain valid. Its name and email address are replaced with a placeholder and its password is removed. Audit log entries about the user keep the user ID and IP address.
  • Deleting a user account also permanently deletes the canvases that user owns.
  • Deleting a canvas or folder in the web client moves it to the user's trash. It is removed when it is deleted from the trash or the trash is emptied.
  • Backups are files you create and manage. Canvus Server does not delete them on a schedule. See Server backup and restore.

Outbound connections

Canvus Server starts these connections itself. Everything else is a connection to the server.

Connection When How to avoid it
MultiTaction license server When an administrator activates a license online (web UI, --activate, or CANVUS_LICENSE_KEY in containers). Use offline activation.
MultiTaction license server (https://ls2026.multitaction.com) Each time the server starts, to fetch the license service's current signing key. The request carries no license key or host ID. If the host cannot be reached, the server starts normally. Allow outbound HTTPS to this host, or leave it blocked: a licensed server keeps working without it.
MultiTaction license server (https://ls2026.multitaction.com) When an administrator clicks Submit usage report for a usage-reported license. Use Export to file and upload the report yourself. See usage-reported licenses.
MultiTaction crash reporting (diagnostics.multitaction.com) When the server process crashes. On by default. See Crash reports.
Your SMTP server When the server sends email, for example to confirm an email address or reset a password. Leave email unconfigured.
Your OpenID Connect provider When single sign-on is set up: to read the provider's configuration and signing keys, and to exchange each sign-in for tokens. Leave single sign-on off.

Installing and updating also downloads software from MultiTaction (container images, installer packages), started by you.

Crash reports

Crash reporting is on by default. When the server process crashes, it writes a crash dump and uploads it to MultiTaction's crash reporting service at diagnostics.multitaction.com, together with the server version and the most recent lines of the server log. A crash dump is a snapshot of the server process at the moment of the crash. The crash dump and the log lines can contain data the server held in memory or logged, including canvas content, email addresses and IP addresses.

  • On Windows, the upload uses HTTPS.
  • On Linux and in containers, the upload uses unencrypted HTTP.

To switch crash reporting off:

  • Containers. Add MULTI_DISABLE_CRASH_REPORTER: "1" to the environment of the canvus service in podman-compose.yml, then recreate the container with sudo podman-compose up -d. Any value switches it off.
  • Linux standalone. Run sudo systemctl edit mt-canvus-server, add the lines below, save, and restart with sudo systemctl restart mt-canvus-server.

    [Service]
    Environment=MULTI_DISABLE_CRASH_REPORTER=1
    
  • Command line. --no-crash-reporter switches it off for that run of mt-canvus-server.

  • Windows standalone. The Windows service does not read --no-crash-reporter or MULTI_DISABLE_CRASH_REPORTER, and has no setting to switch crash reporting off. To stop uploads, block outbound connections from the server to diagnostics.multitaction.com in your firewall or proxy.

Process hardening

  • Containers. The server runs as the unprivileged canvus user. Its only extra capability lets it listen on ports 80 and 443. The database container publishes no port on the host.
  • Linux standalone. The server runs as the canvus system user, which has no login shell. The systemd unit limits its capabilities to binding ports below 1024. The bundled PostgreSQL also runs as canvus and listens on the loopback address only.
  • Windows standalone. The mt-canvus-server and mt-canvus-postgres services run as LocalSystem. The installer restricts the data folder, the configuration file and the database password file to SYSTEM and Administrators. The bundled PostgreSQL listens on the loopback address only.

Signed software

  • The public container images (combined, postgres and media) are signed with Notation. See Verifying image signatures.
  • Release builds of the Windows installer (MSI) and the server programs inside it are code-signed.
  • The Linux packages are published in a signed APT repository at apt.multitaction.com: when you install from it, apt checks the repository's signature (InRelease and Release.gpg) and the package checksums it lists. The .deb files themselves are not individually signed, so a .deb you download on its own carries no signature.