Custom Attack Graph on AttackIQ – Simulating a Full Adversary Workflow
Attack Flow Overview
For this experiment, I built a custom attack graph by chaining multiple existing AttackIQ scenarios into a coherent adversary workflow. The objective was to replicate a realistic attacker kill chain, moving from execution to data exfiltration.
The stages included:
- Execution – mshta.exe execution (LOLBins abuse, T1218.005)
- Defense Evasion – Stop Windows Defender Service (T1562.001)
- Privilege Escalation – UAC Bypass Attempt (T1548.002)
- Credential Access – OS Credential Dumping: LSA Secrets (T1003.004)
- Exfiltration – Simulated HTTP POST exfiltration (T1041)
This illustrates how individual adversary techniques can be combined into a repeatable workflow, rather than tested in isolation.
Why I Designed This Graph
The sequence of techniques in this attack graph was chosen to mirror how an advanced adversary would logically progress through an enterprise environment.
-
Execution (T1218.005 – mshta.exe abuse)
Attackers frequently begin with a LOLBins-based execution path to bypass application controls and leverage trusted binaries. Usingmshta.exe
as the entry point ensures the workflow starts with a realistic, stealthy execution vector. -
Defense Evasion (T1562.001 – Stop Windows Defender Service)
Once initial code execution is established, adversaries often seek to disable endpoint defenses before escalating privileges or persisting. Placing this step early in the chain reflects a common priority: maximizing stealth and reducing detection risk. -
Privilege Escalation (T1548.002 – UAC Bypass Attempt)
With defenses weakened, the next step is to achieve elevated privileges. UAC bypass techniques are a natural escalation path after evasion, enabling access to sensitive system resources and paving the way for credential theft. -
Credential Access (T1003.004 – LSA Secrets)
Once administrative-level execution is possible, adversaries often move to harvest credentials from memory or secure storage. Dumping LSA secrets provides material for lateral movement and persistence, representing a pivotal inflection point in the kill chain. -
Exfiltration (T1041 – Simulated HTTP POST)
The final stage simulates the removal of sensitive data from the environment. Using a standard channel such as HTTP POST reflects realistic attacker behavior, where ordinary protocols are abused to avoid detection.
By chaining these techniques, the attack graph demonstrates a cohesive end-to-end intrusion scenario: stealthy execution → defense evasion → privilege escalation → credential harvesting → data exfiltration. This order reflects both attacker tradecraft and the logical dependencies between techniques.
Alternative Paths Explored
While designing the graph, I considered including:
- Lateral Movement (T1021 – Remote Services) after credential dumping.
- Persistence mechanisms like registry modifications or scheduled tasks.
- Alternative exfiltration channels, such as DNS tunneling or cloud service misuse.
These will be explored in future iterations to expand the complexity of the graph.
Blue Team Perspective
From a defensive design standpoint:
- Each stage of the graph maps cleanly to a MITRE ATT&CK technique.
- This mapping allows defenders to review detection coverage and response playbooks against chained techniques.
- The graph format makes it easier to align security controls with the full attacker workflow, rather than isolated events.
Closing Thoughts
Even though the scenarios were based on existing modules, combining them into a custom adversary workflow provided a meaningful step toward more realistic emulation.