Back to Services
Cloud Security Assessment

Prove Your Cloud
Actually Holds

CSPM tooling tells you what is misconfigured. We tell you which of those misconfigurations chain into a working attack path. AWS, Azure, GCP and OCI — IAM graph analysis, Kubernetes admission testing, serverless + IaC scanning, all mapped to CIS Benchmarks, NIST 800-204D and MITRE ATT&CK Cloud.

AWS · Azure · GCP · OCI
provider coverage
IAM attack graph
every path, per-edge PoC
EKS / AKS / GKE
Kubernetes admission testing
30 days
free retest window

Service overview

An AxVeil cloud security assessment combines posture review with adversary-led validation. We start with a read-only audit role and extract the full configuration of every account, subscription, project or compartment in scope — IAM policies, network paths, storage permissions, KMS / Key Vault usage, compute hardening, container registries, Kubernetes RBAC, serverless triggers, IaC modules. That baseline is compared against the CIS Foundations Benchmark for the target provider, cross-referenced with the AWS Well-Architected Security Pillar, the Microsoft Cloud Adoption Framework Secure baseline, the Google Cloud Architecture Framework Security pillar, the CSA Cloud Controls Matrix v4, and NIST SP 800-204D where the engagement touches the CI/CD pipeline.

What separates an AxVeil engagement from CSPM output is the second half: we model exploitable paths through the configuration. The IBM Cost of a Data Breach Report 2024 attributes 40% of breaches to data stored across multiple environments — cloud-public, cloud-private, on-prem — and the median dwell time for a cloud-resident breach was 277 days. The recurring failure mode is not a single missing checkbox. It is a chain: an overly broad cross-account trust, an EC2 instance role with iam:PassRole, a Lambda with secret material in its environment, a Kubernetes service account bridged to a powerful IAM role through IRSA. CSPM tools flag the individual nodes; the engagement maps the graph.

Engagements scope cleanly to the standard you need to satisfy. SaaS targeting SOC 2 Type 2 typically asks for an AWS or GCP CSPM review plus IAM attack-path mapping on the production account. PCI DSS v4.0 cloud environments ask for segmentation testing across cardholder data environment boundaries and quarterly external coverage. RBI-regulated entities ask for the cyber security framework's annual penetration test extended to cover the cloud control plane. Kubernetes-heavy estates ask for cluster-by-cluster RBAC and admission-controller review against the NSA / CISA Kubernetes Hardening Guidance.

Four sub-offerings

CSPM gap review

Read-only configuration capture across every account / subscription / project / compartment in scope, evaluated against the CIS Foundations Benchmark for the target provider. Cross-referenced with the AWS Well-Architected Security Pillar, Microsoft CAF Secure baseline, Google Cloud Architecture Framework Security pillar and CSA Cloud Controls Matrix v4. Output is a prioritised gap register with each finding tagged to its control, blast radius and remediation owner.

IAM attack-path mapping

Full IAM graph extraction across AWS IAM, Azure RBAC + Entra ID, GCP IAM and OCI IAM. Cross-account / cross-tenant trust relationships, sts:AssumeRole and OIDC federation chains, iam:PassRole edges, lambda:InvokeFunction-as-privesc, condition-key bypass and confused-deputy paths. Output is a directed-graph diagram showing every path from an external or low-trust principal to a sensitive resource, with a PoC command per edge.

Container & Kubernetes testing

EKS, AKS, GKE, OKE and self-managed Kubernetes. RBAC role-binding analysis, admission controller review (Pod Security Standards, OPA / Gatekeeper, Kyverno), service-account-token abuse, network policy validation, container image registry and provenance review, and the IAM bridge between Kubernetes service accounts and cloud roles (IRSA, Workload Identity, Azure AD pod identity). Findings reference the CIS Kubernetes Benchmark v1.9 and the NSA / CISA Kubernetes Hardening Guidance.

Serverless + IaC scanning

Lambda / Azure Functions / Cloud Functions / OCI Functions trigger surface, execution-role privilege review, secrets-in-environment audit, event-injection PoCs across SNS / EventBridge / Pub/Sub / Service Bus. Terraform, CloudFormation, Bicep and Pulumi modules scanned with Checkov, tfsec and KICS, with manual review of high-blast-radius modules (networking, KMS, IAM root policies). CI/CD pipeline review aligns to NIST SP 800-204D where in scope.

Methodology

Seven phases derived from NIST SP 800-204D, the CIS Foundations Benchmarks, MITRE ATT&CK Cloud and the AWS Well-Architected Security Pillar. Multi-cloud estates run the same shape per provider with shared reporting.

01
Scoping & cloud inventory
Define accounts / subscriptions / projects / compartments in scope, target environments (prod vs non-prod), data classifications, Rules of Engagement, read-only audit role provisioning (AWS SecurityAudit + ViewOnlyAccess, Azure Reader + Security Reader, GCP Security Reviewer, OCI Auditor). Written authorisation for any active exploitation against control or data plane.
02
CSPM baseline & configuration review
Automated configuration capture against the CIS Foundations Benchmark for the target provider (AWS v3.0, Azure v2.1, GCP v3.0, OCI v2.0). Outputs cross-referenced with AWS Well-Architected Security Pillar, Azure Cloud Adoption Framework Secure baseline, and CSA Cloud Controls Matrix v4 mappings.
03
Identity & access-path analysis
Full IAM graph extraction. Cross-account trust enumeration, role-assumption chains, condition-key bypass paths, sts:AssumeRole and OIDC-federation abuse, Azure AD conditional-access gaps, GCP service-account impersonation chains, OCI dynamic-group misconfiguration. Privilege-escalation paths surfaced as a directed graph with PoC steps per edge.
04
Workload, container & Kubernetes testing
Compute hardening review (IMDSv2 enforcement, GuestConfig, OS Login). Container registry scanning, image-provenance review, EKS / AKS / GKE / OKE RBAC and admission-controller testing (Pod Security Standards, OPA / Gatekeeper, Kyverno), service-account-token abuse, network policy validation, supply-chain attestation review where SLSA / in-toto in use.
05
Serverless, data & IaC scanning
Lambda / Azure Functions / Cloud Functions / OCI Functions trigger and IAM review, event-injection PoCs, secrets-in-environment audit. S3 / Blob / GCS / Object Storage public-exposure and encryption review. Terraform / CloudFormation / Bicep / Pulumi IaC scanned with Checkov, tfsec and KICS, with manual review of high-blast-radius modules.
06
Segmentation, exploitation & impact mapping
Where in scope, prove the IAM and network paths chain into impact: cross-account pivot, VPC / VNet peering abuse, Transit Gateway / Hub-Spoke segmentation bypass, IMDS-to-role escalation, secret extraction from compromised compute. Findings rated CVSS v3.1 + v4.0 and mapped to MITRE ATT&CK Cloud (Enterprise matrix) TTPs.
07
Reporting & retest
Draft walkthrough call, written report with executive summary, IAM attack-path diagram, full technical findings, remediation per finding with Terraform / Bicep / CLI snippets where appropriate, and one free retest within 30 days. Letter of Attestation issued on PASS.

Frameworks & standards

Every finding is tagged to the control your auditor or platform team will reference.

CIS Foundations Benchmark — AWS v3.0CIS Foundations Benchmark — Azure v2.1CIS Foundations Benchmark — GCP v3.0CIS Foundations Benchmark — OCI v2.0CIS Kubernetes Benchmark v1.9NIST SP 800-204DNIST SP 800-53 Rev 5 (cloud overlay)MITRE ATT&CK Cloud (Enterprise matrix)AWS Well-Architected — Security PillarMicrosoft Cloud Adoption Framework — SecureGoogle Cloud Architecture Framework — SecurityCSA Cloud Controls Matrix v4NSA / CISA Kubernetes Hardening Guidance

Deliverables

A written report (typically 70–140 pages depending on cloud-account count and Kubernetes footprint) plus machine-readable artefacts. Every finding is reproducible from the report alone.

CSPM gap report
Prioritised register of every configuration gap against the CIS Foundations Benchmark for the target provider. Each entry tagged with control reference, affected resource ARN / resource ID, blast radius, remediation owner and effort estimate. Exportable as CSV / JSON for ticketing.
IAM attack-path diagram
Directed-graph diagram (SVG + PDF) showing every modelled path from an external or low-trust principal to a sensitive resource. Each edge annotated with the IAM action / trust relationship / condition-key gap that enables it, plus a PoC command in the report appendix. Re-renderable from the underlying JSON for your own dashboards.
Segmentation testing report
Cross-account / cross-VPC / cross-VNet / cross-project network-path analysis. Where in scope, validated with live probes from each segment. Includes Transit Gateway / Hub-Spoke / VPC Service Controls boundary review and a per-boundary PASS/FAIL.
Kubernetes review
Per-cluster RBAC role-binding inventory, admission-controller posture, service-account-token risk, IRSA / Workload Identity bridge analysis, and CIS Kubernetes Benchmark v1.9 control-by-control PASS/FAIL.
Remediation guidance
Per-finding fix steps written for the cloud-platform engineer, not just the security team. Includes Terraform / Bicep / CLI snippets, IAM policy diffs, condition-key examples and reference links to the provider's own guidance and the relevant NIST / CIS clause.
Free retest within 30 days
One full retest of every Critical, High and Medium finding marked remediated. Retest report appended to the original PDF, attack-path diagram re-rendered post-remediation. Letter of Attestation issued on full PASS.
Sample IAM attack-path edge (JSON export)
{
  "id": "AXV-CLOUD-2026-0007",
  "title": "EC2 instance role can iam:PassRole into AdministratorAccess via Lambda",
  "severity": "Critical",
  "cvss_v3_1": "AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H",
  "cvss_score": 9.6,
  "provider": "AWS",
  "attack_path": [
    "principal: arn:aws:iam::111122223333:role/web-tier-ec2",
    "edge: iam:PassRole -> arn:aws:iam::111122223333:role/lambda-admin",
    "edge: lambda:CreateFunction + lambda:InvokeFunction",
    "target: AdministratorAccess via lambda execution"
  ],
  "mitre_attack_cloud": ["T1078.004", "T1098.001", "T1548.005"],
  "cis_aws": ["1.16", "1.17"],
  "nist_800_53": ["AC-6(1)", "AC-6(2)", "AC-6(7)"],
  "remediation": "Scope iam:PassRole on web-tier-ec2 to an explicit allow-list via iam:PassedToService and a resource ARN list; remove the wildcard.",
  "retest_status": "pending"
}

Engagement timeline

Calendar weeks for a single-provider, 5-to-15-account engagement with one Kubernetes cluster in scope. Multi-cloud estates run providers in parallel under a shared report.

Week 0Scoping call, Rules of Engagement, audit-role provisioning (SecurityAudit / Security Reader / Security Reviewer / Auditor)
Week 1CSPM capture, IAM graph extraction, Kubernetes RBAC inventory, IaC repo clone
Week 2Attack-path modelling, container + admission-controller testing, serverless review
Week 3Segmentation validation, post-exploitation impact mapping, draft report
Week 4Walkthrough call, written report delivered, attack-path diagram handover
Week 4-8Remediation window owned by your platform team
Week 8 (target)Free retest, Letter of Attestation, post-remediation attack-path diagram

Frequently asked questions

Which cloud providers do you cover?+

AWS, Microsoft Azure, Google Cloud Platform and Oracle Cloud Infrastructure. Each engagement is staffed by a tester whose primary specialism is the target provider — AWS engagements lead with Security Audit role analysis, Azure with Defender for Cloud + Entra ID review, GCP with IAM Recommender + Org Policy review, OCI with Cloud Guard + IAM compartment review. Multi-cloud estates are scoped as one engagement with a unified report and a per-provider methodology annex.

What is the difference between CSPM tooling and a cloud security assessment?+

CSPM (Cloud Security Posture Management) tools — Wiz, Prisma Cloud, AWS Security Hub, Defender for Cloud, GCP SCC — continuously flag misconfigurations against a benchmark. They are excellent at the breadth dimension. A cloud security assessment goes beyond that: a tester validates which CSPM findings are exploitable, models real IAM attack paths (which the tooling rarely surfaces with full graph context), tests Kubernetes admission controls, performs container escape research where applicable, and proves segmentation actually holds. The deliverable is exploitable risk, not a list of green and red checks.

How do you map IAM attack paths?+

We pull the full IAM graph using read-only audit roles (AWS SecurityAudit, Azure Security Reader, GCP Security Reviewer, OCI Auditor) and analyse it with PMapper / Cloudsplaining / Principal Mapper-style tooling backed by manual review. Each edge in the graph — iam:PassRole, sts:AssumeRole, lambda:InvokeFunction, role-trust-policy condition-key gaps — is evaluated for exploitability. The deliverable includes a directed-graph diagram showing every path from an external or low-trust principal to a sensitive resource, with PoC commands per edge.

Do you test our Kubernetes clusters?+

Yes. EKS, AKS, GKE, OKE and self-managed Kubernetes are all in scope on request. We test RBAC role bindings, admission controllers (Pod Security Standards, OPA / Gatekeeper, Kyverno), service-account-token abuse paths, network policy enforcement, container-image provenance, and the IAM bridge between Kubernetes service accounts and cloud roles (IRSA, Workload Identity, Azure AD pod identity). Findings reference the NSA / CISA Kubernetes Hardening Guidance and CIS Kubernetes Benchmark v1.9.

Which standards and frameworks do you align to?+

CIS Foundations Benchmarks (AWS v3.0, Azure v2.1, GCP v3.0, OCI v2.0), NIST SP 800-204D (Strategies for the Integration of Software Supply Chain Security in DevSecOps CI/CD Pipelines), MITRE ATT&CK Cloud (Enterprise matrix), AWS Well-Architected Framework Security Pillar, Microsoft Cloud Adoption Framework Secure baseline, Google Cloud Architecture Framework Security pillar, CSA Cloud Controls Matrix v4, and the relevant provider-specific Kubernetes benchmarks. Findings are tagged to the controls your auditor will ask about.

Is the testing safe to run against production?+

Default posture is non-destructive. Configuration capture and IAM graph extraction are read-only. Active testing — IMDS exploitation PoCs, SSRF-to-role-credential chains, container escape PoCs — runs only against pre-agreed targets with rollback procedures documented. Anything that touches data integrity or availability happens in a staging environment unless production testing is explicitly authorised in writing under the Rules of Engagement.

Related work

Scope your cloud security assessment

Send the provider mix, account / subscription count, Kubernetes footprint and audit deadline. We respond with a fixed-fee proposal, lead-tester biography and a sample IAM attack-path diagram from a comparable engagement.

Request a scoping call →