← Back to list
Security & Privacy
#BAS#침해공격시뮬레이션#MITRE ATT&CK#CTEM#AEV
Last updated · 2026-11-10

Breach and Attack Simulation (BAS)

1. Overview

Definition: Breach and Attack Simulation (BAS) is an adversary-centric security validation technology that safely and automatically reproduces a real attacker's tactics, techniques, and procedures (TTP) in order to continuously verify whether security controls — detection, prevention, response — actually work as intended.

Traditional security investment has treated the mere adoption (existence) of a tool — "we bought a firewall," "we deployed EDR" — as evidence of safety. But the fact that a control is installed and the fact that it actually works in the face of a real attack are two entirely different things. This is because control drift — a policy misapplied, a detection rule disabled during an update, logs not forwarded to the SIEM — happens constantly. BAS emerged precisely to close this gap between "is it installed" and "does it actually work."

The background can be organized into three currents. First, the explosion of the attack surface and the pace of change. In environments where cloud, SaaS, and remote work keep assets in constant flux, a penetration test run once or twice a year is already a snapshot of the past. Second, the normalization of detection blind spots. The attacker's average dwell time reported in global incident analyses (Mandiant and others) has recently fallen to around ten days, yet many breaches still go undetected for days to weeks. This suggests EDR and SIEM have many rule gaps where they are "on but cannot see." Third, the demand for automated, continuous validation. Human-led red teaming and penetration testing excel in depth but are expensive and intermittent, unable to keep up with an environment that changes daily and weekly. BAS secures continuity by turning attack scenarios into a library of code and playbooks and running them repeatedly.

Its characteristics can be summarized in four points.

  • Safety (Safe-by-design): Instead of detonating real malware, it runs neutralized payloads and emulation with no operational disruption. It reproduces only the same behavioral patterns, without actually encrypting or exfiltrating data.
  • ATT&CK mapping: By mapping scenarios to the TTPs of the MITRE ATT&CK framework, it shows "which tactical stage is breached" in a common language shared by organizations and vendors.
  • Continuity and automation: Integrated into schedules and CI/CD pipelines, it validates continuously without human intervention and runs repeatedly like a regression test.
  • Evidence-based prioritization: Rather than a plain vulnerability list, it proposes remediation priorities based on "paths that actually succeeded in intrusion or detection evasion."

In short, BAS aims to convert the assumption "we can block the attack" into evidence backed by data. This shift is the starting point of the CTEM and AEV currents discussed later.

2. Conceptual Structure and Operating Principles

A BAS platform consists of four layers, each forming a mutual feedback loop, and this cycle constitutes the core of "continuous validation."

  • Content layer: supplies the latest threats as ATT&CK-based scenarios (the material of validation)
  • Execution layer: reproduces safe payloads via agent, network, or SaaS (performing the attack)
  • Analysis layer: converts control responses into scores and heatmaps (gap identification)
  • Orchestration layer: automatically links remediation and re-validation (closed loop)
graph TD
  subgraph Content["① Content layer(threat intelligence)"]
    TI["Threat intelligence/CTI"]
    LIB["Scenario library(MITRE ATT&CK TTP)"]
  end
  subgraph Exec["② Execution layer"]
    AG["Agent/network/SaaS executor"]
    PL["Safe payload(neutralized emulation)"]
  end
  subgraph Target["③ Target controls(validation objects)"]
    EP["Endpoint EDR/AV"]
    NW["Firewall/IPS/proxy/WAF"]
    SIEM["SIEM/SOAR/SOC"]
  end
  subgraph Analyze["④ Analysis·orchestration"]
    SC["Detection·prevention scoring(gap analysis)"]
    REP["Report/priority/mitigation guide"]
  end
  TI --> LIB --> AG --> PL
  PL -->|"attack reproduction"| EP
  PL -->|"attack reproduction"| NW
  PL -->|"traffic/log trigger"| SIEM
  EP -->|"detect/block result"| SC
  NW -->|"detect/block result"| SC
  SIEM -->|"alert raised?"| SC
  SC --> REP
  REP -->|"rule-improvement feedback"| SIEM

The content layer is the heart that determines BAS quality. It collects the latest threat intelligence (CTI) — ransomware, infostealers, APT campaigns — converts it into attack scenarios, and maps them to ATT&CK tactics (reconnaissance → initial access → privilege escalation → lateral movement → exfiltration). For example, a scenario such as "LockBit-family ransomware behavior" is decomposed into a detailed chain of techniques like phishing-attachment execution (T1566), credential theft (T1003), lateral movement via remote-service abuse (T1021), and data encryption (T1486). Since stale content falls into the trap of "repeatedly validating only attacks already blocked," the vendor's content refresh cadence and speed of incorporating new threats (SLA) becomes a key selection criterion.

The execution layer reproduces scenarios in the real environment. What matters is safety. Instead of actually encrypting files or exfiltrating data, it generates the same system calls, network patterns, and registry changes in a neutralized form, observing only whether controls detect and block them. For instance, exfiltration validation attempts to send a dummy file containing sensitive patterns to a controlled collection server and watches whether the DLP or proxy blocks it.

The analysis layer aggregates whether each control "detected / blocked / did nothing" into a control effectiveness score (prevention·detection score) and visualizes it as a heatmap (coverage map) on the ATT&CK matrix. A red, empty cell is a detection gap, to which concrete mitigation guidance — adding a SIEM correlation rule, hardening an EDR policy — is linked. For example, if an organization validated the top 200 ATT&CK techniques and got blocked 60%, detected-but-not-blocked 25%, and no-response 15%, the most urgent target is the "15% (30 techniques) with neither detection nor prevention." In this way BAS converts security maturity into quantitative metrics such as block rate, detection rate, and mean time to detect (MTTD), enabling measurable reporting to executives such as "the detection rate rose from 71% to 88% before and after the investment."

The orchestration layer closes this feedback loop. It delivers the mitigation guidance derived from analysis as tickets, SOAR playbooks, and SIEM rule changes, and once improvements are applied, it re-runs the same scenario (regression) to confirm the gap was actually filled. The point that this "validate → improve → re-validate" cycle rotates on a daily/weekly basis rather than quarterly or annually is where BAS decisively differs from one-off diagnostics.

Component Core role Symptom on failure
Content (scenarios) Turn the latest TTPs into validatable playbooks Only stale threats validated, new attacks missed
Execution agent Reproduce attacks with safe payloads Coverage bias (endpoint-centric)
Analysis·scoring Identify and prioritize detect/block gaps Trust erosion from noise and false positives
Orchestration Auto-link rule improvement and action Validation only, no linkage to improvement (dead report)

3. Deployment Models and Attack-Lifecycle Validation

Depending on where the executor sits, BAS splits into agent-based, network-based, and cloud (SaaS)-based, and many commercial products blend them. The sequence below shows the flow of a typical "one round of continuous validation."

sequenceDiagram
  participant Ops as Security operations(SOC)
  participant BAS as BAS platform
  participant Env as Target environment(agent)
  participant Ctrl as Security controls(EDR/FW/SIEM)
  Ops->>BAS: Select scenario/schedule(ATT&CK-based)
  BAS->>Env: Deploy safe attack playbook
  Env->>Ctrl: Reproduce attack behavior(safe payload)
  Ctrl-->>Env: Detect/block or no response
  Env-->>BAS: Collect control responses
  BAS->>BAS: Gap analysis·scoring·ATT&CK heatmap
  BAS-->>Ops: Priority report/mitigation guide
  Ops->>Ctrl: Apply rule/policy improvements
  Ops->>BAS: Re-validate(regression test)

Agent-based installs a lightweight agent on endpoints and servers to precisely reproduce host-level behavior (privilege escalation, persistence, lateral movement). Its strength is depth, but its weaknesses are the burden of agent deployment/management and coverage bias (endpoint-centric). Network-based exchanges malicious traffic between virtual nodes placed at both ends of a segment to validate the blocking power of firewalls, IPS, proxies, and NDR. It is strong for data-center internal communication and DMZ-boundary validation. SaaS/cloud-based inspects phishing, web attacks, and data-exfiltration paths from an external attacker's perspective, with low installation burden and suitability for continuous operation. The models are not mutually exclusive; a mature platform combines all three in a hybrid to reproduce an attack chain spanning endpoint, network, and cloud without breaks.

Deployment model Main validation target Strength Limits·caution
Agent-based Host (EDR, priv-esc·lateral movement) Precise host-behavior reproduction Agent-management burden, coverage bias
Network-based Boundary·internal comms (FW·IPS·NDR) Validate segment·DMZ blocking Weak at observing host-internal behavior
SaaS/cloud-based External intrusion·email·web·exfiltration Minimal install, continuous operation Limited at deep internal-path validation

Representative commercial solutions include Cymulate, AttackIQ, SafeBreach, Picus Security, and Mandiant Security Validation; Pentera, which emphasizes convergence with automated penetration testing, and XM Cyber, which specializes in attack-path analysis, form adjacent spaces. On the open-source side, MITRE's CALDERA and Red Canary's Atomic Red Team are widely used, so even budget-constrained organizations can start ATT&CK-based validation. For example, a domestic financial-sector SOC, to supplement one external penetration test per quarter, runs about 50 core ATT&CK techniques with Atomic Red Team automatically once a week to track changes in EDR detection rate.

4. Comparison with Similar Security-Validation Techniques

Rather than replacing existing validation techniques, BAS is a complement that fills the gaps in frequency, automation, and validation target. The difference comes not from a flat feature list but from the question "what are you trying to confirm." A vulnerability scanner asks "is there a hole," penetration testing and red teaming ask "can a human break through," and BAS asks "does our control block/detect a known attack."

Category Validation question Automation/frequency Strength Limit
Vulnerability scan (VM) Is there a hole (CVE) High/continuous Broad asset coverage No exploit/control validation, false positives
Penetration test Is intrusion possible Low/1–2 times a year Proves real intrusion Point-in-time snapshot, costly
Red team Is the objective achievable Very low Stealthy·creative·realistic Narrow scope, scarce, not repeatable
BAS Do the controls work Very high/continuous Continuous·regression·ATT&CK coverage Known-TTP-centric, lacks creativity

As a concrete case, right after a large manufacturer rolled out a new EDR enterprise-wide, an external penetration test rated the controls "good," but two months later continuous BAS validation revealed that the detection rate of a specific lateral-movement technique (remote-service abuse) had plunged. The cause was control drift in which the EDR agent's auto-update failed on part of the server fleet, freezing the detection rules on an old version. BAS, running repeatedly, exposed in just one week a gap that a point-in-time penetration-test snapshot could never have caught. This is the practical power of continuous validation.

Meanwhile, automated pentesting (e.g., Pentera) is adjacent to BAS but different in grain. Where BAS emphasizes "validating (did it block?) control responses with pre-defined scenarios," automated pentesting autonomously explores "whether a discovered vulnerability can actually be chain-exploited to reach domain-admin privileges." The former proves detection coverage, the latter proves an exploitable path. Recently, as the two approaches converge into AEV, there is a clear trend of putting validation and exploitation in one platform, as with "BAS 2.0."

The key implication is BAS's unique value of "continuity of detection·prevention validation." Even if a penetration test blocks a path it once breached, whether a patch or policy change reopened it a month later cannot be known until the next penetration test. BAS re-runs the same scenario like a regression test and immediately reveals "whether what was supposedly fixed broke again." On the other hand, because BAS centers on a library of known TTPs, creative attacks such as zero-days, social engineering, and physical intrusion remain the province of the human red team. Thus a mature organization layers BAS (continuous, broad) + penetration testing (periodic, deep) + red team (occasional, real) + purple team (collaborative improvement). Purple-team automation — immediately sharing a detection gap found by BAS with the SOC to fix the rule — is especially effective in practice.

5. Deep Dive — Convergence into AEV·CTEM and Exam Outlook

The most important recent trend is that BAS is converging from a standalone category into "Adversarial Exposure Validation (AEV)." In its 2024 Hype Cycle for Security Operations, Gartner consolidated BAS, automated penetration testing, and (automated) red teaming into the higher-level category of AEV; it is reported to have published an AEV Market Guide in March 2025 and the first Magic Quadrant for Exposure Assessment Platforms in November 2025 (since the details are based on vendor and industry summaries, firm assertions are avoided). AEV is defined as the offense-oriented technology group that "mimics attacker TTPs to filter out only exposures that are actually exploitable and validate control effectiveness."

This convergence meshes with the CTEM (Continuous Threat Exposure Management) framework. Among CTEM's five stages (scoping → discovery → prioritization → validation → mobilization), BAS/AEV takes charge of exactly the validation stage. That is, when a vulnerability scanner pours out "exposure candidates," BAS filters out only "those that actually breach controls" and provides the basis for remediation priority. This turns a hollow number like "tens of thousands of CVEs" into an actionable list of "the top N paths that actually succeeded in intrusion." Some vendors, including Pentera, call this "BAS 2.0," combining validation with real exploitation. This current directly connects to this site's [[ctem]], [[mitre-attack]], [[ai-red-teaming]], and [[cyber-threat-intelligence]] topics.

BAS's usefulness is also growing in the domestic context. The "check on the operational effectiveness of security controls" required by certification schemes such as [[isms-p]], and the periodic drill requirements of the financial sector, can be reinforced with objective evidence from BAS's continuous-validation reports. That said, pre-authorization and scope control must be codified so that automated attack behavior is not mistaken for a real breach, and it is realistic to position BAS as a continuous complement rather than a full replacement for outsourced penetration testing.

From a Professional Engineer's perspective, the expected exam directions are threefold. First, the comparative-essay type: "explain the concept, components, and operating principles of BAS and compare it with penetration testing and vulnerability assessment." Second, the integrated-design type: "how will you position BAS and design its operational process when adopting CTEM." Third, the considerations type, asking about "the limits and governance (safety, noise, staffing) of automated security validation." An answer is differentiated by centering on ATT&CK mapping, continuous validation, gap analysis, and purple-team linkage, and placing the latest AEV/CTEM currents in the conclusion.

6. Considerations and Implications

BAS grants the powerful value of "automated validation," but its adoption must be designed on the premise of the following trade-offs.

  • Controlling safety and operational impact: Although BAS claims neutralization, because it triggers real attack behavior, EDR alert floods, automatic isolation, and false-positive blocks can occur. Start in a non-production (staging) and restricted scope at first, and block operational impact and "confusion with a real attack" via whitelists, maintenance windows, and prior SOC notice.
  • Dependence on content freshness: BAS's value is proportional to the currency of its scenario library. Stale content gives a false sense of security by "repeatedly validating only attacks already blocked." Therefore, in vendor selection, make the new-threat-incorporation SLA, CTI integration, and custom-scenario authoring capability key criteria.
  • Building the "validate → improve" closed loop: If reports merely pile up without leading to rule improvement, it becomes a dead dashboard. Linking BAS results to SOAR, tickets, and SIEM rule changes, and automating re-validation (regression) after improvement, is what yields return on investment. This is the essence of purple-team automation.
  • Staffing·maturity·irreplaceable domains: BAS presumes the SOC maturity to perform analysis, tuning, and action. If a low-maturity organization adopts it, alerts only increase while action fails to follow. Also, because BAS is known-TTP-centric, the logical abuse of zero-days, social engineering, physical security, and cloud misconfigurations must be complemented by the human red team and the [[ctem]] discovery stage.
  • Strategic linkage (outlook): Over the medium to long term, BAS is likely to be absorbed as a feature of AEV·exposure-management platforms rather than a standalone product. On adoption, position it as a validation engine within the CTEM architecture rather than a single BAS, and design it to share data with vulnerability management and threat hunting using [[mitre-attack]] coverage as a common metric — this is sustainable.

References


In one line: BAS safely and automatically reproduces an attacker's TTPs to continuously validate the real detection·prevention power of security controls; centered on MITRE ATT&CK coverage and gap analysis, it complements penetration testing and vulnerability assessment and converges into CTEM's validation stage (AEV).