implementing-threat-modeling-with-mitre-attack

使用 MITRE ATT&CK 实施威胁建模

使用时机

在以下情况下使用此技能：

• SOC 团队需要评估针对相关威胁行为者及其 TTP 的检测覆盖情况
• 安全负责人需要根据威胁情报确定防御优先级
• 新环境（云迁移、OT 集成）需要规划检测策略
• 紫队演练需要基于威胁模型开展结构化的对手模拟
• 年度风险评估需要基于 ATT&CK 分析威胁态势

请勿将其作为一次性练习使用——随着对手 TTP 的演变和组织攻击面变化，必须持续更新威胁模型。

前提条件

• 了解 MITRE ATT&CK 框架（Enterprise、ICS、Mobile 或 Cloud 矩阵）
• ATT&CK Navigator 工具（网页版或本地版），用于可视化图层
• 已映射到 ATT&CK 技术 ID 的当前检测规则清单
• 关于攻击你所在行业的对手团体的威胁情报
• 包含关键性分类的组织资产清单

工作流

步骤 1：识别相关威胁行为者

使用 MITRE ATT&CK Groups 研究攻击你所在行业的对手团体：

import requests
import json

# Download ATT&CK STIX data
response = requests.get(
    "https://raw.githubusercontent.com/mitre/cti/master/enterprise-attack/enterprise-attack.json"
)
attack_data = response.json()

# Extract groups and their techniques
groups = {}
for obj in attack_data["objects"]:
    if obj["type"] == "intrusion-set":
        group_name = obj["name"]
        aliases = obj.get("aliases", [])
        description = obj.get("description", "")
        groups[group_name] = {
            "aliases": aliases,
            "description": description[:200],
            "techniques": []
        }

# Map techniques to groups via relationships
relationships = [obj for obj in attack_data["objects"] if obj["type"] == "relationship"]
techniques = {obj["id"]: obj for obj in attack_data["objects"]
              if obj["type"] == "attack-pattern"}

for rel in relationships:
    if rel["relationship_type"] == "uses":
        source = rel["source_ref"]
        target = rel["target_ref"]
        for group_name, group_data in groups.items():
            if source == group_data.get("id") and target in techniques:
                tech = techniques[target]
                ext_refs = tech.get("external_references", [])
                for ref in ext_refs:
                    if ref.get("source_name") == "mitre-attack":
                        group_data["techniques"].append(ref["external_id"])

# Example: Financial sector threat actors
financial_actors = ["FIN7", "FIN8", "Carbanak", "APT38", "Lazarus Group"]
for actor in financial_actors:
    if actor in groups:
        print(f"{actor}: {len(groups[actor]['techniques'])} techniques")
        print(f"  Top techniques: {groups[actor]['techniques'][:10]}")

步骤 2：构建威胁行为者 TTP 概况

为优先级较高的威胁行为者创建 ATT&CK Navigator 图层：

import json

def create_attack_layer(actor_name, techniques, color="#ff6666"):
    """Generate ATT&CK Navigator JSON layer for a threat actor"""
    layer = {
        "name": f"{actor_name} TTP Profile",
        "versions": {
            "attack": "15",
            "navigator": "5.0",
            "layer": "4.5"
        },
        "domain": "enterprise-attack",
        "description": f"Techniques associated with {actor_name}",
        "techniques": [
            {
                "techniqueID": tech_id,
                "tactic": "",
                "color": color,
                "comment": f"Used by {actor_name}",
                "enabled": True,
                "score": 1
            }
            for tech_id in techniques
        ],
        "gradient": {
            "colors": ["#ffffff", color],
            "minValue": 0,
            "maxValue": 1
        }
    }
    return layer

# Create layers for top threat actors
fin7_techniques = ["T1566.001", "T1059.001", "T1053.005", "T1547.001",
                    "T1078", "T1021.001", "T1003", "T1071.001", "T1041"]
layer = create_attack_layer("FIN7", fin7_techniques, "#ff6666")

with open("fin7_layer.json", "w") as f:
    json.dump(layer, f, indent=2)

步骤 3：映射当前检测覆盖情况

导出已映射到 ATT&CK 的当前检测规则：

--- Extract ATT&CK technique mappings from Splunk ES correlation searches
| rest /services/saved/searches
  splunk_server=local
| where match(title, "^(COR|ESCU|RBA):")
| eval techniques = if(isnotnull(action.correlationsearch.annotations),
                       spath(action.correlationsearch.annotations, "mitre_attack"),
                       "unmapped")
| stats count by techniques
| mvexpand techniques
| stats count by techniques
| rename techniques AS technique_id, count AS rule_count

创建检测覆盖图层：

def create_coverage_layer(detection_rules):
    """Generate coverage layer from detection rule inventory"""
    technique_counts = {}
    for rule in detection_rules:
        for tech in rule.get("techniques", []):
            technique_counts[tech] = technique_counts.get(tech, 0) + 1

    layer = {
        "name": "SOC Detection Coverage",
        "versions": {"attack": "15", "navigator": "5.0", "layer": "4.5"},
        "domain": "enterprise-attack",
        "techniques": [
            {
                "techniqueID": tech_id,
                "color": "#31a354" if count >= 2 else "#a1d99b" if count == 1 else "",
                "score": count,
                "comment": f"{count} detection rule(s)"
            }
            for tech_id, count in technique_counts.items()
        ],
        "gradient": {
            "colors": ["#ffffff", "#a1d99b", "#31a354"],
            "minValue": 0,
            "maxValue": 3
        }
    }
    return layer

步骤 4：执行差距分析

将威胁行为者的 TTP 与检测覆盖情况进行叠加比较：

def gap_analysis(threat_techniques, covered_techniques):
    """Identify detection gaps for specific threat actor"""
    gaps = set(threat_techniques) - set(covered_techniques)
    covered = set(threat_techniques) & set(covered_techniques)

    print(f"Threat Actor Techniques: {len(threat_techniques)}")
    print(f"Detected: {len(covered)} ({len(covered)/len(threat_techniques)*100:.0f}%)")
    print(f"Gaps: {len(gaps)} ({len(gaps)/len(threat_techniques)*100:.0f}%)")

    # Prioritize gaps by kill chain phase
    priority_order = {
        "TA0001": 1, "TA0002": 2, "TA0003": 3, "TA0004": 4,
        "TA0005": 5, "TA0006": 6, "TA0007": 7, "TA0008": 8,
        "TA0009": 9, "TA0010": 10, "TA0011": 11, "TA0040": 12
    }

    gap_details = []
    for tech_id in gaps:
        gap_details.append({
            "technique": tech_id,
            "priority": "HIGH" if tech_id.split(".")[0] in ["T1003", "T1021", "T1059"] else "MEDIUM",
            "recommendation": f"Build detection for {tech_id}"
        })

    return {
        "total_actor_techniques": len(threat_techniques),
        "covered": len(covered),
        "gaps": len(gaps),
        "coverage_pct": round(len(covered)/len(threat_techniques)*100, 1),
        "gap_details": sorted(gap_details, key=lambda x: x["priority"])
    }

# Run analysis
result = gap_analysis(fin7_techniques, current_coverage)

步骤 5：制定优先级整改计划

构建检测工程路线图：

threat_model_remediation_plan:
  assessed_date: 2024-03-15
  primary_threats:
    - FIN7 (Financial sector)
    - APT38 (DPRK financial)
    - Lazarus Group (Destructive)

  current_coverage: 64%
  target_coverage: 80%

  priority_1_gaps: # 30-day target
    - technique: T1021.002
      name: SMB/Windows Admin Shares
      data_source: Windows Security Event 5140
      effort: Low
      detection_approach: Monitor admin share access from non-admin workstations

    - technique: T1003.006
      name: DCSync
      data_source: Windows Security Event 4662
      effort: Medium
      detection_approach: Detect DS-Replication-Get-Changes from non-DC sources

  priority_2_gaps: # 60-day target
    - technique: T1055
      name: Process Injection
      data_source: Sysmon EventCode 8, 10
      effort: High
      detection_approach: Monitor cross-process memory access patterns

    - technique: T1071.001
      name: Web Protocols (C2)
      data_source: Proxy/Firewall logs
      effort: Medium
      detection_approach: Detect beaconing patterns in HTTP/S traffic

  priority_3_gaps: # 90-day target
    - technique: T1070.004
      name: File Deletion
      data_source: Sysmon EventCode 23
      effort: Low
      detection_approach: Monitor mass file deletion in sensitive directories

步骤 6：通过对手模拟进行验证

使用 MITRE Caldera 或 Atomic Red Team 测试覆盖情况：

# Using Atomic Red Team to validate coverage for FIN7 techniques
# T1566.001 — Spearphishing Attachment
Invoke-AtomicTest T1566.001

# T1059.001 — PowerShell
Invoke-AtomicTest T1059.001 -TestNumbers 1,2,3

# T1053.005 — Scheduled Task
Invoke-AtomicTest T1053.005

# T1547.001 — Registry Run Keys
Invoke-AtomicTest T1547.001

# T1003 — Credential Dumping
Invoke-AtomicTest T1003 -TestNumbers 1,2

# Verify detections
# Check SIEM for corresponding alerts within 15 minutes

记录模拟结果，以验证威胁模型的准确性。

核心概念

术语	定义
MITRE ATT&CK	基于现实观察构建的对手战术、技术和流程知识库
TTP	战术、技术和流程——对手团体的行为模式
ATT&CK Navigator	用于将 ATT&CK 矩阵可视化为分层热力图的 Web 工具，可展示覆盖情况或威胁概况
差距分析	将威胁行为者的 TTP 与检测覆盖情况进行比较，以识别盲区的过程
威胁知情防御	根据真实对手行为而非理论风险确定防御优先级的安全策略
对手模拟	对威胁行为者的 TTP 进行受控模拟，以验证检测和响应能力

工具和系统

• MITRE ATT&CK Navigator：基于 Web 的可视化工具，用于创建和叠加 ATT&CK 技术图层
• MITRE Caldera：自动化对手模拟平台，可大规模测试检测覆盖情况
• Atomic Red Team：用于验证安全控制的 ATT&CK 技术开源测试库
• CTID ATT&CK Workbench：MITRE 工具，可结合组织情境自定义 ATT&CK 知识库
• Tidal Cyber：基于 ATT&CK 框架制定威胁知情防御计划的商业平台

常见场景

• 年度威胁评估：将排名前 5 的威胁行为者映射到 ATT&CK，与检测覆盖情况叠加比较，并生成差距分析
• 云迁移规划：建模云环境特定威胁（T1078.004、T1537），并规划检测覆盖情况
• 并购安全评估：针对相关威胁行为者，对被收购公司的环境进行威胁建模
• 预算论证：利用差距分析展示需要投入工具的检测盲区
• 紫队规划：根据威胁模型中优先级最高的差距，选择对手模拟场景

输出格式

THREAT MODEL ASSESSMENT — Financial Services Division
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Date:             2024-03-15
Threat Actors:    FIN7, APT38, Lazarus Group
Techniques Total: 87 unique techniques across all actors

DETECTION COVERAGE:
  Covered:     56/87 (64%)
  Gaps:        31/87 (36%)

  Tactic Coverage Breakdown:
    Initial Access:      78%  ████████░░
    Execution:           82%  █████████░
    Persistence:         71%  ████████░░
    Priv Escalation:     65%  ███████░░░
    Defense Evasion:     52%  ██████░░░░  <-- Priority gap
    Credential Access:   58%  ██████░░░░  <-- Priority gap
    Discovery:           45%  █████░░░░░
    Lateral Movement:    61%  ███████░░░
    Collection:          50%  ██████░░░░
    Exfiltration:        55%  ██████░░░░
    C2:                  67%  ███████░░░

TOP PRIORITY GAPS (30-day remediation):
  1. T1055 Process Injection — used by all 3 actors, 0 detections
  2. T1003.006 DCSync — used by FIN7 and Lazarus, 0 detections
  3. T1070.004 File Deletion — evidence destruction, 0 detections

INVESTMENT RECOMMENDATION:
  Closing top 10 gaps requires: 2 detection engineer FTEs, 60 days
  Expected coverage improvement: 64% -> 76%