From Threat Modelling to Threat Detection: Making STRIDE Work in Cloud Environments
One of the biggest mistakes in cloud security is treating threat modelling as a one-off exercise. A diagram is created, risks are identified, a document is produced, and then it stops. The real value of threat modelling is not in the exercise itself but in how it drives detection, visibility, and response.
22 Apr 2026
~10 min read
Saleem Yousaf
The Problem
In many organisations, threat modelling is performed early in the project lifecycle, documented for compliance or assurance, and rarely revisited once systems go live. This creates a gap. You may understand the risks but have no way of knowing when those risks are being exploited in real time.
The Shift: From Design to Detection
To make threat modelling useful, it needs to answer: how would we detect this in real time? This is where STRIDE becomes far more powerful. Instead of just identifying threats, we map them directly to observable behaviours and detection use cases.
// STRIDE threat hunting in cloud landing zones
STRIDE Mapped to Cloud Detection
Spoofing
Becomes suspicious sign-ins, unusual identity provider behaviour, and unexpected role assumptions. Detection sources: Azure AD and Entra ID logs, AWS CloudTrail AssumeRole events, and identity protection alerts.
Tampering
Becomes policy drift, security control changes, and network rule modifications. Detection sources: Azure Policy, AWS Config, activity logs, and infrastructure drift detection.
Information Disclosure
Becomes unusual data access, bulk downloads, and public exposure of storage. Detection sources: S3 and Blob access logs, Defender for Cloud and Macie, and data exfiltration patterns.
Privilege Escalation
Becomes risky role assignments, permission changes, and elevated actions following role changes. Detection sources: IAM logs, Privileged Identity Management, and CloudTrail and Activity Logs.
Denial of Service
Becomes traffic spikes, resource exhaustion, and API abuse patterns. Detection sources: WAF logs, metrics and monitoring alerts, and load balancer telemetry.
Repudiation
Becomes unattributed privileged actions and missing or incomplete audit trails. Detection sources: centralised logging via SIEM and identity and session tracking.
Why This Matters for Cloud Landing Zones
A landing zone is not just about networks, identity, and resource structure. It is the foundation for visibility and detection. If detection is not designed into the landing zone, logs may be incomplete, signals may be missed, and incidents may go undetected.
Architecture matters. But architecture without visibility is blind. The real value comes from connecting design to threats to detection to response.