The Novee Gym is open for business at Black Hat Europe

Book a meeting

The Novee Gym is open for business at Black Hat Europe

Book a meeting

Glossary › Adversary Emulation

Adversary Emulation

Explore Article +

Key Takeaways

  • Adversary emulation recreates how a specific, real threat actor operates to test whether defenses would stop it.
  • An adversary emulation plan turns threat intelligence into an ordered set of TTPs (tactics, techniques and procedures).
  • Mapping every action to MITRE ATT&CK makes results measurable and comparable over time.
  • Breach and attack simulation platforms automate emulation so teams can run it continuously.

What Is Adversary Emulation?

Adversary emulation is a security testing approach in which a team reproduces the known behavior of a specific threat actor, such as a ransomware gang or a state-sponsored group, against its own environment. The goal is to answer a concrete question: if this attacker targeted us, would we prevent, detect and respond to it?

Unlike a generic vulnerability scan or penetration test, emulation follows the attacker’s real playbook, step by step, based on threat intelligence. That makes it a test of people, processes and detection tooling, not just of exposed weaknesses.

Building an Adversary Emulation Plan From Threat Intelligence

  1. Choose the adversary. Pick a group that targets your industry, region or technology stack, using intelligence reports and sources like MITRE ATT&CK’s group pages.
  2. Gather intelligence. Collect what is known about its initial access methods, tools, persistence and objectives.
  3. Extract TTPs. Break the reporting into discrete behaviors, such as spearphishing with macros or credential dumping from LSASS.
  4. Build the adversary emulation plan. Order the TTPs into a realistic attack flow, define the tools or scripts for each step and set success criteria.
  5. Agree scope and safety controls. Set rules of engagement, approved hosts and a stop procedure.
  6. Execute and record. Run each step and log what was blocked, detected or missed.

Public resources speed this up. The Center for Threat-Informed Defense publishes ready-made plans for groups including APT29, FIN6 and menuPass in its Adversary Emulation Library.

Mapping Emulated Behavior to MITRE ATT&CK

MITRE ATT&CK is a public knowledge base of adversary tactics, such as Initial Access or Lateral Movement, and the techniques used to achieve them. Teams tag each emulated step with its ATT&CK technique ID, for example T1059 for Command and Scripting Interpreter.

After the exercise, results are plotted on the ATT&CK matrix, often with the free ATT&CK Navigator. Each technique is marked as prevented, detected, logged only or missed. The heatmap shows defensive gaps at a glance, guides detection engineering and gives a common language for red, blue and leadership teams.

Adversary Emulation vs Adversary Simulation

The terms are often used loosely, but a useful distinction is:

  • Adversary emulation copies a specific named actor’s TTPs as closely as possible. It fits when a known threat targets your sector and you want to validate defenses against it.
  • Adversary simulation imitates realistic attacker behavior in general, without tying every step to one group. It fits broader testing of how defenses handle common attack patterns.

Many programs use both: simulation for coverage of common techniques, emulation for focused tests against the most relevant threats. Both differ from a red team exercise, which tests whether a team can achieve an objective by any realistic means.

Automating Adversary Emulation With Breach and Attack Simulation

Manual emulation takes weeks of planning and expert time, so most teams run it once or twice a year. Breach and attack simulation (BAS) platforms package TTPs as safe, repeatable test cases that run on agents across endpoints, networks and cloud. Open-source options include MITRE Caldera and Red Canary’s Atomic Red Team.

Automation lets teams rerun scenarios after every change, catch detection drift and measure improvement over time. This matches Gartner’s push toward continuous offensive security testing. AI is extending it further: Novee’s research on teaching a small model to hack like a real attacker shows how agents can chain attacker reasoning, not just replay scripts. Three independent agents validate each Novee finding, and Agentic Remediation retests fixes.

FAQ

Yes. MITRE ATT&CK includes matrices for cloud platforms such as AWS, Azure, Google Cloud and Microsoft 365, plus containers. Teams emulate cloud-focused groups by testing identity abuse, misconfigured storage, API key theft and control-plane persistence. Cloud provider rules usually allow testing your own tenants, but check their penetration testing policies first.

Run a full manual emulation at least once a year, and after major changes like a merger, new security tooling or a cloud migration. Automated emulation through breach and attack simulation can run weekly or continuously to catch detection drift. New intelligence about a group targeting your sector is also a good trigger.

It can be, with controls. Teams use benign payloads, avoid destructive actions, agree rules of engagement and keep a stop procedure ready. Many start in staging, then move to production for realistic results. BAS platforms are designed for production with safe test cases, but high-risk steps still need approval and monitoring.

Common choices are well-documented groups with rich public reporting: APT29, APT28, FIN6, FIN7, Carbanak, Lazarus, Sandworm and menuPass, plus ransomware operators such as Conti and LockBit affiliates. The Center for Threat-Informed Defense library and MITRE ATT&CK Evaluations provide ready plans. The best choice depends on which groups target your industry.

It helps, but it is not required. Mature teams get the most from full manual emulation with custom tooling. Smaller teams can start with published emulation plans, Atomic Red Team tests or an automated platform, focusing on a few high-priority techniques. The key prerequisite is logging good enough to see whether each step was detected.