Defending the Nation’s Infrastructure

Executive summary

The threat to United States critical infrastructure has changed character. Adversaries are no longer only stealing data. They are quietly pre-positioning inside the systems that keep the lights on, the water running and benefits flowing, waiting for a moment of their choosing. Defending against that mission requires a different posture from perimeter defense: assume breach, make lateral movement expensive, and build systems that keep delivering essential services through an attack.

This paper sets out the defensive practices that actually blunt today’s campaigns, shows where machine intelligence is a genuine force multiplier for defenders and where it is equally the adversary’s tool, and offers a phased roadmap operators can execute now.

The adversary has a new mission

The defining shift of the current era is pre-positioning. PRC-linked groups such as Volt Typhoon, Salt Typhoon and Silk Typhoon have demonstrated the ability to gain and hold long-term access to communications, energy, water and wastewater, and transportation networks by exploiting compromised credentials, unpatched edge devices and trusted third-party access.

United States intelligence assesses that Volt Typhoon is positioning itself to enable disruptive or destructive attacks against critical infrastructure in a future crisis or conflict, rather than to conduct traditional espionage. There is no public confirmation that its persistence has been fully eradicated from those networks. Salt Typhoon, by contrast, is an espionage operation, and has been reported to have infiltrated at least nine United States telecommunications carriers.

The common thread is stealth. These actors blend in by using legitimate native system tools, a technique known as living off the land, and valid credentials. They have been observed testing access to operational-technology assets using default vendor credentials. There is often no malware to catch, only an authorized-looking session doing unauthorized things.

Why critical infrastructure is uniquely exposed

Three structural realities make these systems harder to defend than a typical enterprise.

IT and OT convergence

Intrusions typically begin in IT networks and then pivot into the operational-technology environments that actually control physical processes: a substation, a pump, a switch. The boundary between the office network and the plant floor is where the damage potential lives.

Legacy and long-lived systems

OT equipment runs for decades, often cannot be patched on an enterprise cadence, and was never designed for an adversarial internet. Replacement cycles are measured in capital programs, not sprints.

Trusted third parties and exposed edge

Vendor remote access and internet-facing edge devices are repeatedly the initial foothold, and the long dwell times these campaigns achieve mean a breach can sit undetected for months before anything is noticed.

The defensive foundation

Guidance converging across CISA, NSA and NIST points in one direction: resilience over prevention alone. The assumption is no longer whether an adversary gets in, but how little they can accomplish once they do.

Adopt zero trust, starting with assume breach

Federal guidance on hardening OT with zero trust calls for enforcing micro-segmentation, hardening controllers and improving visibility in order to disrupt the ICS cyber kill chain, operating on the assumption that the environment is already compromised. Every request is verified and no network location is inherently trusted.

Segment and isolate aggressively

This is the single highest-leverage move against pre-positioned threats. CISA’s CI Fortify initiative urges operators to be able to deliver essential services under emergency conditions, turning off third-party and business-network connections to an OT network and operating in isolation for weeks to months when facing an active threat. The capability to disconnect and keep running has to exist before it is needed.

Fix the entry points the campaigns actually use

• Patch internet-facing edge devices on a priority cadence, separate from the general patch queue.
• Eliminate default and shared credentials, including on OT assets where vendor defaults are routinely left in place.
• Enforce phishing-resistant multifactor authentication.
• Tightly control and time-box vendor remote access.

Because these actors rely on stolen credentials and native tooling, identity hygiene and behavioral monitoring matter more than signature-based antivirus.

Instrument for detection of legitimate-looking activity

Monitor for lateral-movement tooling and anomalous use of native administrative utilities across OT VLANs. The goal is to catch a valid credential being used in an invalid way, which is a question about behaviour rather than about signatures.

Map it to a governing framework

NIST Cybersecurity Framework 2.0 gives operators a shared language: Govern, Identify, Protect, Detect, Respond and Recover. The addition of Govern in version 2.0 puts accountability and risk ownership at the center, which is exactly where critical-infrastructure programs tend to be weakest.

Machine intelligence as a force multiplier, for both sides

Super intelligence has moved from pilot to production in security operations, and the honest framing is that it cuts both ways.

On offense, the adversary is already scaling

The 2026 CrowdStrike Global Threat Report found that AI-enabled adversary operations increased 89 percent year over year, and a 2026 Dark Reading reader poll found 48 per cent of cybersecurity professionals now rank agentic attacks as their top threat vector, ahead of deepfakes and supply-chain compromise. Attacks that once needed a team now need one operator and an agent.

On defense, it is how operators keep pace

The practical wins are concrete: faster threat detection, automation of a large share of routine analyst work, and response at machine speed through behavioral anomaly detection, alert triage and correlation, and bounded automated containment. Detection and behavioral analysis can be layered over existing industrial systems without ripping out and replacing legacy OT, which matters a great deal when the equipment cannot be touched.

Autonomy must be bounded and governed

This is where discipline separates a credible program from a risky one. NIST is developing an AI Risk Management Framework profile for trustworthy intelligence in critical infrastructure, launched in April 2026, and the work is ongoing rather than final. Evaluation methods for agentic security are still maturing and will not catch every failure mode.

A practical implementation roadmap

The Varmoda approach

War Mode is the right mindset for this problem: defend as though the adversary is already inside, because for critical infrastructure assume-breach is no longer a posture, it is the documented reality.

Varmoda builds critical-infrastructure defense the way the guidance now demands. Zero-trust architecture and aggressive segmentation come first. Resilience is engineered so that essential services survive isolation. Super intelligence is deployed as a governed force multiplier rather than an unsupervised actor. Varmoda pairs cleared public-sector delivery with a Microsoft-anchored security practice and sequences autonomy against the authorization boundary and the mission, not against the hype, as set out in the companion paper on agentic workflows inside GCC High.

Conclusion

The nation’s infrastructure is being contested quietly, right now, by adversaries whose goal is leverage in a future crisis. The defense is not a product, it is a discipline: verify everything, make movement expensive, assume breach, build to operate through an attack, and let machine intelligence accelerate defenders while humans stay accountable for the decisions that matter. Operators who adopt that posture deny the adversary the one thing pre-positioning depends on, which is time inside the walls, unseen.

← Back to Insights