DedSec Project

Cloud Logging, Detection and Cross-Cloud Investigation — DedSec Project

Cloud Logging, Detection and Cross-Cloud Investigation Cloud investigations depend on control-plane and identity evidence distributed across services. Learn normalized event models…

Cloud investigations depend on control-plane and identity evidence distributed across services. Learn normalized event models, immutable collection, correlation, time, multi-account structure and investigation playbooks.

Learning objectives

By the end of this module you should be able to:

Deep concepts

1. control-plane audit logs

Cloud audit logs record administrative/API actions such as identity, policy, networking, storage, and resource changes. Enable and centralize them before an incident; retroactive investigation is impossible for events the provider/account never retained.

2. identity and token context

A cloud event is meaningful only when caller identity, assumed role/service principal, session/token context, source, target resource, and organization/account/tenant are correlated. Normalize temporary identities back to their parent workload or human where possible.

3. data-plane versus management-plane logs

Management-plane logs describe configuration/control actions, while data-plane logs describe access to workloads or stored data. High-value investigations often require both because a policy change and subsequent data access happen in different telemetry systems.

4. central aggregation

Send logs to a security account/project or independent store with narrow write/admin permissions. Cross-account aggregation reduces the chance that compromise of one workload lets an attacker erase the only copy of its control-plane evidence.

5. integrity and retention

Use provider/object controls, immutability where appropriate, retention policy, export verification, and restricted deletion to preserve evidence. Retention should reflect incident-detection latency, legal/privacy requirements, and cost rather than a single arbitrary number.

6. time synchronization and ordering

Distributed cloud events may arrive late, use different timestamps, or represent server/client time differently. Preserve original timestamps and ingestion time, correlate with stable request/session IDs, and avoid assuming displayed order equals causal order.

7. cross-cloud normalization

AWS, Azure, GCP, SaaS, and identity providers use different names for principals, resources, actions, and outcomes. Normalize into a common investigation schema while preserving provider-specific raw fields so analysts can pivot without losing semantics.

8. investigation pivots and evidence preservation

Start from a known indicator—identity, resource, IP, request ID, key, or time window—and pivot across identity, control plane, workload, network, and data access. Export only necessary evidence, hash important artifacts, and document query/time-zone assumptions for reproducibility.

Authorized lab sequence

Knowledge checks

Answer these without looking at the notes:

Next modules

Recommended related modules: 019, 023, 037, 047, 059, 076, 080, 105, 106. From the main menu, choose Search lessons to find related sections across the full guide.