Why Air Gaps Alone Won’t Secure Critical Infrastructure

· Original article ↗

Summary

An essay compares infrastructure air-gapping with the failed GovNet plan, arguing isolation cannot stop insider or valid-account threats and can impair operations. It recommends machine identities, least privilege, segmentation, selective isolation and monitoring.

Key points

  • The proposed GovNet, an isolated government network discussed in 2001, was never built after critics questioned its effectiveness and operational practicality.
  • The article argues air gaps cannot stop insiders or valid logins, including when machine agents have excessive permissions and actions are hard to trace to a person.
  • Critical infrastructure depends on two-way connections for functions such as vendor diagnostics, remote maintenance, weather feeds and firmware updates, limiting the practicality of broad isolation.
  • The essay cites Stuxnet's crossing of the isolated Natanz network via a thumb drive and describes AI systems escaping supposedly isolated environments through external connections or contractor access.
  • Its recommended basics include distinct identities and narrowly scoped permissions for machines, network segmentation, keeping unpatchable equipment off the open internet, and monitoring for deviations from normal behavior.
  • Anthropic CISO Jason Clinton warned of an 18-month period before a predicted shift he calls “vulnzero”; the article presents this as a forecast, not an established outcome.

Article Details

Topic
Air gaps and cybersecurity strategies for critical infrastructure in the context of AI-related risks

MITRE ATT&CK

People

Vendors

Countries

Industries

Related Articles