Back to insights
Cybersecurity

Securing Remote Mine Sites Over Satellite Links: A Checklist

A remote shaft on a slow satellite link needs the same protection as head office. This checklist covers endpoint protection, patching, backup and remote access that work over constrained links.

Most security tooling is designed for an office with fibre. A mine is not an office. Our mining customers run head offices in Johannesburg, processing plants hundreds of kilometres away and shafts reached by VSAT satellite or microwave links of between 2 and 10 Mbps, with latency that can exceed 600 ms. The Minerals Council South Africa reported 472 000 people employed in the industry in 2025; many of them work at sites like these.

A control that stops working when the link drops, or saturates the link when it is up, is not a control. This checklist sets out what we check at each remote site so that protection, patching, backup and application access hold up on the connectivity the site actually has.

Measure the link before you design for it

Every other decision depends on three numbers per site: usable bandwidth in each direction, round-trip latency and how often the link is down. Satellite links are often asymmetric, with far less upload than download, and backup and telemetry both travel upstream.

  • Record bandwidth, latency and monthly outage time for each site, measured from the site.
  • Note what else shares the link: fleet management, safety telemetry and SCADA remote access usually take priority over IT traffic.
  • Agree a bandwidth budget for security traffic per site, and design everything below to fit it.

Protect: endpoint protection that does not need the cloud

An antivirus that waits for a signature update or a cloud verdict is exposed for as long as the link is down. SentinelOne runs its static and behavioural AI engines on the device. SentinelOne states that it does not rely on connectivity to protect the device, and that a disconnected endpoint detects malware locally.

  • Deploy the agent to every Windows-based device at the site, including control-room terminals and engineering laptops that leave the mine.
  • Set the policy so that detection and rollback act automatically.
  • Check which devices have not reported for longer than the site's usual outage window. A silent agent is either on a failed link or has been removed, and both need a visit. Our remote-worker endpoint guide covers the same principle for staff who move between sites.

Detect: monitoring and patching inside a bandwidth budget

A Windows cumulative update downloaded by forty devices over one 4 Mbps link will fail or crowd out production traffic. N-able N-central handles this with a probe on the site network that caches patches and installers once; devices at the site fetch them from the probe rather than over the WAN. N-able's documentation describes the cache as reducing bandwidth usage and improving download times; it is enabled per site.

  • Place a Windows probe at each site, with caching enabled, on a machine that stays on.
  • Schedule patch downloads to the probe for the quietest hours and the installs for the shift change that production allows.
  • Alert on the link itself. A failing radio or satellite terminal usually shows in monitoring data before a user calls.
  • Keep agents on devices that cannot be patched, such as older HMIs tied to a vendor release, and mark them for isolation and replacement. The IT/OT segmentation checklist explains where those devices belong on the network.

Recover: backup that fits the upstream link

Backing up a site server over a satellite link has two problems: the first full backup, and the restore. Cove Data Protection addresses both. Its change-tracking keeps incremental backups small, and the Backup Manager can limit bandwidth by time of day in Kbit/s or Mbit/s, so the nightly backup stays within the budget agreed above. Our TrueDelta article covers the data reduction.

For the restore, Cove's LocalSpeedVault keeps an additional copy on a local disk or SMB share at the site. Backups go to the vault and the cloud at the same time; a restore comes from the vault first and does not depend on the internet connection. If the vault is unavailable or out of sync, the restore comes from the cloud automatically. The cloud copy remains the immutable one; the vault is what makes a site restore practical over a slow link.

  • Seed the first full backup at a planned window, within the daily budget.
  • Size each site's LocalSpeedVault at two to three times the data it protects, which is N-able's recommendation.
  • Include the files that live only at the site: mine planning models, survey data, PLC programs and HMI projects.
  • Test a restore from the vault and one from the cloud at each site, and record the time each took.

Remote access: deliver the application, not the data

Geologists and engineers at a shaft need SAP, Deswik or Datamine, and copying a multi-gigabyte model over satellite is not practical. Citrix virtual apps and desktops keep the application and its data in the data centre and send only the screen to the site.

Citrix's Enlightened Data Transport (EDT) is a UDP-based protocol that Citrix describes as giving a better experience on challenging long-haul connections and more throughput on unreliable networks. In the default Preferred mode, the client tries EDT and TCP in parallel, falls back to TCP if UDP is blocked and retries EDT every five minutes.

  • Allow UDP as well as TCP on port 443 to the NetScaler Gateway, or sessions quietly fall back to TCP and the satellite latency shows.
  • Leave EDT MTU discovery on; it is enabled by default and avoids the fragmentation that makes a slow link worse.

Put it in writing for the auditors

The Mine Health and Safety Act, POPIA and King IV all ask, in different words, whether the controls work everywhere the business operates. The evidence is the per-site record from the steps above: the measured link, the agents reporting, the patch state, the last successful backup and restore, and the transport protocol in use. For managed clients we keep that record and review it with each site.

Sources

Running sites the fibre never reached? Talk to OAS about managed security for mining.

Keep exploring

All insights