Real-world cloud security. Explained and applied.
Technical postmortem of the Storm-2949 case in Azure: how an identity-based chain and abuse of App Services enabled exfiltration of Key Vault secrets and SQL Database data without deploying malware, and which signals and controls failed in corporate environments.
Dirty Frag (CVE-2026-43284) allows, under specific conditions, escalating to root on the AWS EKS node from a compromised container. This article describes the attack surface, operational signals, and how to apply seccomp profiles and validate them in EKS to break the exploitation chain.
Session cookie theft via Adversary-in-the-Middle (AitM) allows an attacker to reuse an already authenticated session in the cloud console, even when the administrator has MFA. This article breaks down operational signals, real business impact, and practical mitigations based on conditional access, network validation, and reducing session surface area.
The supply chain in container registries fails when we treat public base images as “trusted” and when we use mutable tags like :latest. This article grounds the abuse scenario, why a corporate registry is not synonymous with security, and how to operate ECR/ACR/GCR with immutability, push-time scanning, and signature policies (Cosign/Notation) that actually block unsafe deployments.
An attacker with read permissions can avoid a massive SELECT and exfiltrate complete databases by sharing snapshots with an external account or making them temporarily public. This operational guide explains how RDS/EBS and equivalents are abused, which events to look for in logs, and how to block it with guardrails (SCP/Azure Policy) and real-time alerts.
“We encrypt it with KMS” is not a security measure if the default key does not enforce additional controls. Provider-managed encryption often allows implicit decryption for anyone with permissions on the resource. This article debunks the myth and explains how to restrict key usage (CMK) with policies and conditions such as kms:ViaService, and what changes between AWS, Azure, and GCP.
In RAG architectures, a PDF can become a “control channel” for your agent: it injects instructions the model obeys and, if the runtime has broad permissions (S3/DB/email), the real consequence is data and credential exfiltration. This article breaks down the abuse, the architectural failure, and how to harden permissions and action flows without breaking the product.
In analytics and ML it is common to dump datasets into “temporary” buckets to speed up experiments. When those buckets become public (read or, worse, write), the risk goes from PII leakage to data manipulation and infrastructure abuse. This operational guide explains how these blind spots are created, how to detect them, and how to close the exposure without breaking pipelines.
A malicious package.json or setup does not need to exploit your application: it just needs to run during install in CI/CD and talk to the metadata service to steal temporary credentials. This article breaks down the scenario, the signals, and how to put guardrails in place so your pipelines do not download packages from unverified sources.
Egress in serverless often ends up “open by default”. If a function is compromised, that path is used to exfiltrate data or contact C2. This operational guide explains how to restrict real destinations (VPC endpoints, NAT, firewalls and DNS), how to apply guardrails, and how to verify that outbound control works in production.