
Detecting Multi Stage Attacks on AWS
A practical introduction to how security teams can correlate activity across AWS services to identify suspicious attack sequences that may not be obvious from a single alert.
A security alert rarely tells the complete story.
An attacker may perform several actions that individually look normal. For example, an identity may make an API request from a new location, access a resource it has not used before, and then make changes to another service.
Looking at each event separately may not reveal an attack. Looking at the sequence can reveal the bigger picture.
This is the idea behind detecting multi-stage attacks.
Why Individual Alerts Are Not Always Enough
Consider this simplified scenario:
- An unfamiliar identity accesses an AWS environment.
- The identity calls an API that it has never used before.
- The identity assumes another role.
- A resource is modified.
- Data is accessed shortly afterward.
Each action might have a legitimate explanation.
However, when these events happen close together and involve the same identity or related resources, they become much more interesting from a security perspective.
The important question is therefore not only:
"Did something suspicious happen?"
It is also:
"What happened before and after it?"
CloudTrail: Understanding the Activity
AWS CloudTrail provides a record of AWS API activity.
When investigating suspicious activity, useful information can include:
- The identity that performed an action
- The API operation
- The source IP address
- The time of the event
- The AWS Region
- The affected resource
This information provides the timeline needed to connect multiple events.
For example, instead of investigating a single unusual API call, a security team can examine the events surrounding that call.
GuardDuty: Detecting Suspicious Activity
Amazon GuardDuty continuously analyzes relevant AWS data sources to help identify potential threats.
GuardDuty findings can provide important security signals.
However, a finding should not always be treated as the complete incident.
A useful investigation asks:
- What identity generated the activity?
- What resources were accessed?
- Were there other unusual actions around the same time?
- Did the activity continue after the finding?
- Does the behavior match the identity's normal activity?
This turns individual findings into an investigation timeline.
CloudWatch: Bringing Signals Together
Amazon CloudWatch can help collect, analyze, and monitor operational and security-related signals.
When security events are correlated with application and infrastructure information, teams can build better context around an incident.
For example:
- An unusual authentication event
- followed by an unexpected API call
- followed by a resource modification
- followed by unusual network activity
The sequence is much more meaningful than any single event.
Think in Attack Sequences
A useful way to approach cloud security is to think in terms of sequences rather than isolated events.
For example:
Stage 1: Initial Access
An attacker obtains or abuses credentials.
Stage 2: Discovery
The attacker begins discovering available resources, identities, or permissions.
Stage 3: Privilege Expansion
The attacker attempts to obtain additional permissions or assume another role.
Stage 4: Resource Access
The attacker accesses valuable resources or sensitive data.
Stage 5: Impact
The attacker modifies, deletes, encrypts, or exfiltrates resources.
Not every attack follows this exact pattern, but thinking in stages helps security teams investigate activity more effectively.
A Simple Investigation Approach
When an unusual AWS event appears, start with five questions:
1. Who performed the action?
Identify the IAM principal, role, or other identity involved.
2. Where did it originate?
Look at the source address, Region, and other available request information.
3. What happened immediately before it?
Search for related activity from the same identity or resources.
4. What happened afterward?
Look for additional API calls, permission changes, resource modifications, or data access.
5. Is the sequence normal?
Compare the activity with the expected behavior of the workload or identity.
This approach helps separate an isolated anomaly from a potentially coordinated attack.
Security Is About Context
One of the most important lessons in cloud security is that context matters.
An unusual API call does not automatically mean that an account has been compromised.
Similarly, a normal API call does not automatically mean that everything is safe.
Security teams need to consider:
- Identity
- Timing
- Resource
- Location
- Previous activity
- Subsequent activity
- Expected behavior
When these signals are connected, security teams can make better decisions.
Final Takeaway
AWS security should not be viewed as a collection of independent alerts.
CloudTrail can provide the activity timeline.
GuardDuty can provide threat detection signals.
CloudWatch can help with monitoring and correlation.
The real value comes from connecting these signals and understanding the sequence of events.
When investigating an AWS security event, don't stop at:
"What happened?"
Also ask:
"What happened before it, what happened after it, and do these events make sense together?"
That shift from isolated alerts to connected activity can significantly improve cloud security investigations.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article