DeepTeam is a framework to red team LLMs and AI agents.
import pytest
from deepteam.frameworks import MITRE
from deepteam.red_teamer.risk_assessment import RiskAssessment
from deepteam.vulnerabilities import BaseVulnerability
from deepteam.attacks import BaseAttack
from deepteam.frameworks.risk_category import RiskCategory
from deepteam.frameworks.mitre.risk_categories import MITRE_CATEGORIES
from deepteam import red_team
class TestMITRE:
def test_mitre_init(self):
"""Test that MITRE ATLAS framework can be instantiated."""
framework = MITRE()
assert framework is not None
def test_mitre_name(self):
"""Test that MITRE ATLAS framework has correct name."""
framework = MITRE()
assert framework.name == framework.get_name() == "MITRE ATLAS"
def test_mitre_description(self):
"""Test that MITRE ATLAS framework has correct description."""
framework = MITRE()
assert (
framework.description
== "A structured knowledge base of adversarial tactics, techniques, and procedures (TTPs) used against AI and ML systems. Extends MITRE ATT&CK® principles to the AI threat surface, testing across six adversarial tactics: reconnaissance (information gathering), resource development (capability building), initial access (entry point exploitation), ML attack staging (model-specific preparation), exfiltration (data/model theft), and impact (manipulation/degradation)."
)
def test_mitre_default_categories(self):
"""Test that all six MITRE categories are included by default."""
framework = MITRE()
assert set(framework.categories) == {
"reconnaissance",
"resource_development",
"initial_access",
"ml_attack_staging",
"exfiltration",
"impact",
}
def test_mitre_partial_categories(self):
"""Test that MITRE can be initialized with limited categories."""
framework = MITRE(categories=["reconnaissance", "impact"])
assert set(framework.categories) == {"reconnaissance", "impact"}
def test_mitre_vulnerabilities_exist(self):
"""Test that MITRE framework defines vulnerabilities."""
for risk_category in MITRE_CATEGORIES:
assert hasattr(risk_category, "vulnerabilities")
assert len(risk_category.vulnerabilities) > 0
def test_mitre_vulnerabilities_are_instances(self):
"""Test that all vulnerabilities are instances of BaseVulnerability."""
for risk_category in MITRE_CATEGORIES:
for vuln in risk_category.vulnerabilities:
assert isinstance(vuln, BaseVulnerability)
def test_mitre_attacks_exist(self):
"""Test that MITRE framework defines attacks."""
for risk_category in MITRE_CATEGORIES:
assert hasattr(risk_category, "attacks")
assert len(risk_category.attacks) > 0
def test_mitre_attacks_are_instances(self):
"""Test that all attacks are instances of BaseAttack."""
for risk_category in MITRE_CATEGORIES:
for attack in risk_category.attacks:
assert isinstance(attack, BaseAttack)
def test_mitre_category_vulnerability_mapping(self):
"""Test that all categories map to vulnerabilities properly."""
categories = MITRE_CATEGORIES
assert set([category.name for category in categories]) == {
"reconnaissance",
"resource_development",
"initial_access",
"ml_attack_staging",
"exfiltration",
"impact",
}
for risk_category in categories:
assert isinstance(risk_category, RiskCategory)
assert all(
isinstance(v, BaseVulnerability)
for v in risk_category.vulnerabilities
)
def test_mitre_category_attack_mapping(self):
"""Test that all categories map to attacks properly."""
categories = MITRE_CATEGORIES
assert set([category.name for category in categories]) == {
"reconnaissance",
"resource_development",
"initial_access",
"ml_attack_staging",
"exfiltration",
"impact",
}
for risk_category in categories:
assert isinstance(risk_category, RiskCategory)
assert all(isinstance(v, BaseAttack) for v in risk_category.attacks)
def test_mitre_vulnerability_names_present(self):
"""Test that key MITRE vulnerabilities are present."""
vuln_names = []
for risk_category in MITRE_CATEGORIES:
vuln_names.extend(
[v.__class__.__name__ for v in risk_category.vulnerabilities]
)
expected_vulns = [
"IllegalActivity",
"PromptLeakage",
"PIILeakage",
"ExcessiveAgency",
"IntellectualProperty",
"Competition",
"GraphicContent",
"RBAC",
"SSRF",
"DebugAccess",
"ShellInjection",
"SQLInjection",
"CustomVulnerability",
"RecursiveHijacking",
]
for name in expected_vulns:
assert (
name in vuln_names
), f"Expected vulnerability {name} not found"
def test_mitre_attack_names_present(self):
"""Test that key MITRE attacks are present."""
attack_names = []
for risk_category in MITRE_CATEGORIES:
attack_names.extend(
[a.__class__.__name__ for a in risk_category.attacks]
)
expected_attacks = [
"Roleplay",
"PromptInjection",
"LinearJailbreaking",
"TreeJailbreaking",
"SequentialJailbreak",
"CrescendoJailbreaking",
"Leetspeak",
"ROT13",
"PromptProbing",
]
for name in expected_attacks:
assert name in attack_names, f"Expected attack {name} not found"
def test_mitre_attack_weights_valid(self):
"""Test that all attacks have valid weights."""
for risk_category in MITRE_CATEGORIES:
for attack in risk_category.attacks:
assert hasattr(attack, "weight")
assert isinstance(attack.weight, int)
assert 1 <= attack.weight <= 3
def test_mitre_framework_with_red_team(self):
import random
"""Test MITRE ATLAS integration with red_team pipeline."""
def mock_model_callback(prompt: str, turns=None) -> str:
return "This request violates internal policy."
categories = [
"reconnaissance",
"resource_development",
"initial_access",
"ml_attack_staging",
"exfiltration",
"impact",
]
random_category = random.choice(categories)
risk_assessment = red_team(
model_callback=mock_model_callback,
framework=MITRE(categories=[random_category]),
async_mode=False,
ignore_errors=False,
)
assert isinstance(risk_assessment, RiskAssessment)
assert risk_assessment is not None