Skip to main content

Security and architecture

An overview for a security assessment of UCS: architecture, infrastructure, sign-in, permissions, logging, updates and backup. Details are on the linked pages.

Architecture​

UCS is a contact centre and IP PBX, run in the cloud or on-premise. It runs on a Linux virtual server and consists of server components and the web-based Operator application.

It is usually run as an HA solution in an active/passive cluster with real-time database replication. It can also run on a single server, with availability provided by the virtualisation platform.

Infrastructure requirements​

  • A virtual server on VMware with full vCPU and RAM reservation – UCS is a real-time system, see UCS installation and VMware prerequisites.
  • Ubuntu or RHEL. The current and the previous release are always supported (now Ubuntu 24.04 and 26.04, RHEL 9 and 10); for new installations we recommend Ubuntu 26.04 or RHEL 10.

Network communication​

The minimum to run:

PortProtocolService
443tcpHTTPS
5060 / 5061udp / tcpSIP / SIP over TLS
3478udpSTUN/ICE
10000–20000udpRTP / SRTP

IP phones also need DHCP, TFTP, NTP and other services. When integrated with Active Directory, UCS opens outbound Kerberos (88) and LDAPS (636) connections.

The full list of ports is on the Communication page.

Server hardening​

UCS is designed to run without administrator (root) privileges:

  • All UCS services run as the unprivileged system user ucs, and systemd removes all kernel privileges (capabilities) from them. The services cannot gain higher privileges, change kernel parameters, load kernel modules or change the system time.
  • Changes to the system configuration (network, firewall, TLS certificate, DHCP, NTP, SMTP, VPN) are made by a single separate service, ucs-root. It offers a fixed set of actions with validated parameters; UCS never passes it paths, commands or whole configuration files. Only the ucs user may call it, every request is written to the system log, and the service itself runs in a systemd sandbox – with no network access and write access only to the files it manages. It validates new configuration before using it and rolls back on failure.
  • UCS installs no sudo or polkit rules and no setuid programs.
  • Internal services listen on the local address only and are reachable from the network only through the Traefik reverse proxy. Only the telephony services (SIP, RTP, SIPREC) and the services for IP phones listen on the network directly, see Communication.
  • Configuration files with passwords and keys are readable only by the service that uses them. The database and ARI passwords are generated randomly during installation.

In transit:

  • The web administration, the Operator application and the API are available over HTTPS only. The services for IP phones are also available over HTTP, for older phones.
  • SIP signalling to IP phones and trunks can run over TLS (port 5061) and voice over SRTP. It is enabled per device and per trunk. All TLS versions supported by Asterisk are supported; in practice TLS 1.2 and 1.3, as older versions are disabled by the OS cryptographic policy.
  • Web softphones connect over encrypted WebSocket (WSS) and carry voice over DTLS-SRTP.

At rest: data on disk (database, recordings) is protected by disk or virtual server encryption at the infrastructure level.

Authentication, SSO and MFA​

  • SSO via Kerberos (SPNEGO/GSSAPI), see Settings for SSO, or via OpenID Connect. If SSO fails, the sign-in form is offered.
  • MFA is available when signing in via OpenID Connect with Microsoft Entra ID (Microsoft Online), where it is enforced by the identity provider's policies.
  • The mobile app can require biometric authentication (FaceID) each time it is opened, see Mobile application.

Active Directory, LDAP and identity providers​

UCS synchronises users from Active Directory over LDAP. Permissions are assigned by group membership, including nested groups.

Provisioning via SCIM 2.0 is supported as an alternative, see User provisioning. In a hybrid environment (on-premise AD and Entra ID), SCIM provisioning from Entra ID requires a Microsoft Entra ID P1 licence for every synchronised user.

Users, roles and permissions​

Access control is role-based (RBAC): each user has exactly one role, which determines their permissions. Access rights (ACLs) are set on a tree of groups and are inherited by subgroups.

Audit and security logging​

The audit log is a native part of the system; users with the Access auditlog permission can read it. Application logs are written to the systemd journal and can also be written to files in parallel.

Both the audit log and the application logs can be exported in real time to an external system via syslog or OpenTelemetry, see Observability.

Updates and security vulnerabilities​

UCS is installed and updated with the operating system's standard package manager (apt or dnf). OS packages are not pinned, so security updates can be installed as Canonical or Red Hat release them.

We recommend updating regularly on a 14-day cycle. We can provide updates as a service, on-premise installations included.

Processed and stored data​

UCS stores in its database the configuration and user accounts, call detail records (CDR), statistics, contact-centre e-mails and the audit log. Outside the database it stores call recordings and, if used, their AI transcriptions.

The data retention period is configurable.

AI services (call transcription, voicebot) always use the customer's own account with Azure or another provider. The data is therefore processed by the provider under the customer's contract, in the region the customer chooses.

Supplier access​

INSOFT service access to on-premise installations goes through VPN and SSH with authentication by security tokens.

Backup and recovery​

Backup works on three levels:

  • a backup of the whole virtual server,
  • a regular database dump, by default daily with 30-day retention (see ucs-db-dump),
  • an incremental backup of call recordings.
Single serverHA cluster
RPO (possible data loss)the database dump interval, 24 h by default, can be reduced to e.g. 1 h0 – the database is replicated in real time
RTO (recovery time)1 to 4 hours, depending on system sizetakeover within 1 minute