Skip to content
Home / Services / Cloud Pentest
Pillar I · Offensive Security

Cloud Pentest AWS, Azure & GCP

A cloud pentest is a penetration test of your cloud environment — AWS, Azure and GCP — that exploits IAM flaws, storage exposure, misconfigurations, CI/CD pipeline secrets and containers/Kubernetes to prove, with evidence, what an attacker could reach starting from your account.

What a cloud pentest is

A cloud pentest assesses the security of an environment hosted on providers such as AWS, Azure and Google Cloud from a real attacker's perspective. Unlike a traditional network pentest, the focus is on the cloud control plane — identities, permissions, policies, keys and how managed services are configured — not just the operating systems and applications running inside it.

Under the shared responsibility model, the provider secures the infrastructure, but the configuration of identities, buckets, networks and pipelines belongs to the customer. That is exactly the layer where most cloud incidents happen. Berghem combines automated enumeration with manual exploitation to turn insecure configurations into concrete attack scenarios, with a reproducible escalation path and impact-prioritized recommendations.

For regulated institutions, cloud scope must also meet sector requirements — including the BACEN/CMN resolutions on cybersecurity and cloud computing. We structure the test to produce evidence useful to your compliance process.

What's included

•AWS, Azure and Google Cloud (GCP)
•IAM — users, roles, policies and escalation
•Storage exposure (S3, Blob, GCS)
•Managed-service misconfigurations
•Secrets in CI/CD and Infrastructure as Code
•Containers, registries and Kubernetes
•Report with PoC, evidence and remediation
Attack surface

What we test in a cloud pentest

We cover both the cloud control plane and the data plane. Each domain below is enumerated, manually validated and chained into attack paths that demonstrate real impact — never just an isolated tool finding.

IAM · Identity

IAM and privilege escalation

We analyze users, roles, groups and policies for excessive permissions, overly permissive policies, misconfigured cross-account trust and escalation paths. We map how a low-privilege credential can reach account administrator by abusing actions such as iam:PassRole, role assumption and poorly designed federation.

Storage · Data

Storage exposure

Public buckets and blobs with loose ACLs are a leading cause of cloud leaks. We check S3, Azure Blob Storage and Google Cloud Storage for anonymous access, undue listing, public policies, unprotected versioning and exposed sensitive data, including forgotten snapshots and backups.

Config · Services

Service misconfigurations

Misconfigured managed services open the door for attackers. We review open security groups and firewalls, internet-facing databases, serverless functions with broad permissions, poorly managed KMS keys, disabled logging and instance metadata exposed to SSRF.

CI/CD · Secrets

Secrets in CI/CD and IaC

Pipelines and repositories often store access keys, tokens and credentials in variables, logs and Infrastructure as Code files. We hunt for leaked secrets, over-privileged runners and pipeline service accounts, and misconfigured OIDC — vectors that jump straight into production cloud.

K8s · Containers

Containers and Kubernetes

We assess vulnerable images, exposed registries, permissive RBAC, privileged pods, service accounts with mounted tokens and container escape paths. In managed clusters (EKS, AKS, GKE) we check the link between Kubernetes identity and cloud IAM, a common escalation vector.

Network · Edge

Network and external surface

We enumerate the external surface exposed by the cloud — endpoints, load balancers, management APIs and published assets — and test segmentation, VPCs, peering and routes. The goal is to identify what is reachable from the internet and how it connects initial access to the rest of the environment.

Providers

Coverage by platform

Each provider has its own identity model and managed services. We adapt the methodology to the specifics of each cloud, whether a single environment or multi-cloud.

Amazon Web Services

AWS pentest

AWS penetration testing focused on IAM, STS and role assumption, S3 buckets, Lambda functions, EC2 metadata (IMDS), EKS and KMS keys. We map escalation paths and lateral movement across accounts of an AWS Organization.

Microsoft Azure

Azure pentest

Assessment of Azure and Entra ID (Azure AD): roles and RBAC, service principals and managed identities, Azure Blob Storage, Key Vault, Functions and AKS. We analyze the boundaries between the identity tenant and resource subscriptions.

Google Cloud

GCP pentest

Penetration testing in Google Cloud focused on IAM and service accounts, impersonation and token generation, Cloud Storage, Cloud Functions, GKE and Secret Manager. We check permission inheritance across the organization, folder and project hierarchy.

Methodology

How we run the test

We work with agreed rules of engagement and a secure channel with the client from start to finish. The test can start from an external perspective (no credentials) or assume a low-privilege identity to measure the blast radius of a compromised credential.

01 · Scope

Scope and rules of engagement

We define in-scope accounts, subscriptions and projects, the testing model (external or assumed credential), windows and limits. We align authorization with the provider where applicable and agree on the exercise objectives.

02 · Enumeration

Enumeration and mapping

We inventory identities, permissions, services and exposed assets using enumeration tooling and manual review. We build the graph of relationships between identities and resources to locate the most promising escalation paths.

03 · Exploitation

Exploitation and escalation

We exploit insecure configurations in a controlled way — privilege escalation, data access, lateral movement across services and accounts — chaining flaws into full attack scenarios while preserving the stability of the environment.

04 · Report

Report and remediation

We deliver each finding with a proof of concept, evidence, severity rating and impact-prioritized remediation. We present the results technically and offer a retest to validate the fixes applied.

FAQ

Frequently asked questions about cloud pentesting

What's the difference between a cloud pentest and a traditional network pentest?
A network pentest focuses on operating systems, services and segmentation. A cloud pentest focuses on the control plane — identities, permissions, policies and managed services like storage, serverless functions and Kubernetes — where most incidents in AWS, Azure and GCP occur.
Do I need cloud-provider authorization for the test?
Most providers allow testing on the customer's own resources without prior approval, but certain activities require notice. We check AWS, Azure and GCP testing policies at project start and align scope and authorization before any action.
Is the test run externally or with cloud credentials?
Both are possible. The external perspective measures what an attacker with no access can reach. The assumed perspective — with a low-privilege identity — measures the blast radius of a compromised credential, a more realistic scenario to assess IAM escalation.
Does a cloud pentest meet compliance requirements?
Yes. We structure the test and report to produce evidence useful to frameworks such as PCI-DSS and ISO 27001 and to the BACEN/CMN resolutions on cybersecurity and cloud, with findings, severity and remediation documented in an auditable way.

Ready to test your cloud?

In a 30-minute conversation we map your AWS, Azure or GCP environment and design the ideal cloud-pentest scope.

Talk to a specialist