feat(phase-28): add scoring system, operational metrics and executive dashboard (T-224 to T-226)
This commit is contained in:
468
backend/app/services/operational_metrics_service.py
Normal file
468
backend/app/services/operational_metrics_service.py
Normal file
@@ -0,0 +1,468 @@
|
||||
"""Operational metrics service — MTTD, MTTR, Detection Efficacy, and more.
|
||||
|
||||
Calculates security operations KPIs from test data and audit logs.
|
||||
"""
|
||||
|
||||
from datetime import datetime, timedelta
|
||||
from typing import Optional
|
||||
|
||||
from sqlalchemy import func, case, and_, or_, extract
|
||||
from sqlalchemy.orm import Session
|
||||
|
||||
from app.models.test import Test
|
||||
from app.models.technique import Technique
|
||||
from app.models.test_detection_result import TestDetectionResult
|
||||
from app.models.audit import AuditLog
|
||||
from app.models.enums import TestState, TestResult
|
||||
|
||||
|
||||
def _safe_stats(values: list[float]) -> dict:
|
||||
"""Compute mean, median, min, max from a list of floats."""
|
||||
if not values:
|
||||
return None
|
||||
sorted_vals = sorted(values)
|
||||
n = len(sorted_vals)
|
||||
return {
|
||||
"mean_hours": round(sum(sorted_vals) / n, 1),
|
||||
"median_hours": round(sorted_vals[n // 2], 1),
|
||||
"min_hours": round(sorted_vals[0], 1),
|
||||
"max_hours": round(sorted_vals[-1], 1),
|
||||
"sample_size": n,
|
||||
}
|
||||
|
||||
|
||||
# ── MTTD (Mean Time to Detect) ───────────────────────────────────────
|
||||
|
||||
|
||||
def calculate_mttd(db: Session) -> Optional[dict]:
|
||||
"""Calculate Mean Time to Detect.
|
||||
|
||||
For each validated test: time between entering red_executing and
|
||||
entering blue_evaluating (extracted from audit_log timestamps).
|
||||
"""
|
||||
# Get validated tests that have both timestamps available
|
||||
# Using audit log entries for state transitions
|
||||
tests = (
|
||||
db.query(Test)
|
||||
.filter(Test.state == TestState.validated)
|
||||
.all()
|
||||
)
|
||||
|
||||
detection_times = []
|
||||
for test in tests:
|
||||
# Find the red_executing and blue_evaluating transition timestamps
|
||||
red_start = (
|
||||
db.query(AuditLog.timestamp)
|
||||
.filter(
|
||||
AuditLog.entity_type == "test",
|
||||
AuditLog.entity_id == str(test.id),
|
||||
AuditLog.action.in_(["test_start_execution", "start_execution"]),
|
||||
)
|
||||
.order_by(AuditLog.timestamp.asc())
|
||||
.first()
|
||||
)
|
||||
|
||||
blue_start = (
|
||||
db.query(AuditLog.timestamp)
|
||||
.filter(
|
||||
AuditLog.entity_type == "test",
|
||||
AuditLog.entity_id == str(test.id),
|
||||
AuditLog.action.in_(["test_submit_red", "submit_red"]),
|
||||
)
|
||||
.order_by(AuditLog.timestamp.asc())
|
||||
.first()
|
||||
)
|
||||
|
||||
if red_start and blue_start and blue_start[0] > red_start[0]:
|
||||
hours = (blue_start[0] - red_start[0]).total_seconds() / 3600
|
||||
detection_times.append(hours)
|
||||
|
||||
return _safe_stats(detection_times)
|
||||
|
||||
|
||||
# ── MTTR (Mean Time to Respond/Remediate) ─────────────────────────────
|
||||
|
||||
|
||||
def calculate_mttr(db: Session) -> Optional[dict]:
|
||||
"""Calculate Mean Time to Respond.
|
||||
|
||||
For tests with remediation_status = completed: time between
|
||||
detection_result being set and remediation_status = completed.
|
||||
"""
|
||||
# Tests with completed remediation
|
||||
tests = (
|
||||
db.query(Test)
|
||||
.filter(
|
||||
Test.remediation_status == "completed",
|
||||
Test.blue_validated_at.isnot(None),
|
||||
)
|
||||
.all()
|
||||
)
|
||||
|
||||
response_times = []
|
||||
for test in tests:
|
||||
# Find when remediation was completed from audit log
|
||||
remediation_complete = (
|
||||
db.query(AuditLog.timestamp)
|
||||
.filter(
|
||||
AuditLog.entity_type == "test",
|
||||
AuditLog.entity_id == str(test.id),
|
||||
AuditLog.action.ilike("%remediation%"),
|
||||
)
|
||||
.order_by(AuditLog.timestamp.desc())
|
||||
.first()
|
||||
)
|
||||
|
||||
detection_time = test.blue_validated_at
|
||||
if remediation_complete and detection_time:
|
||||
hours = (remediation_complete[0] - detection_time).total_seconds() / 3600
|
||||
if hours > 0:
|
||||
response_times.append(hours)
|
||||
|
||||
return _safe_stats(response_times)
|
||||
|
||||
|
||||
# ── Detection Efficacy ───────────────────────────────────────────────
|
||||
|
||||
|
||||
def calculate_detection_efficacy(db: Session) -> dict:
|
||||
"""Calculate detection efficacy: detected / total validated tests."""
|
||||
validated_tests = (
|
||||
db.query(Test)
|
||||
.filter(Test.state == TestState.validated)
|
||||
.all()
|
||||
)
|
||||
|
||||
total = len(validated_tests)
|
||||
if total == 0:
|
||||
return {
|
||||
"percentage": 0,
|
||||
"detected": 0,
|
||||
"partially": 0,
|
||||
"not_detected": 0,
|
||||
"total": 0,
|
||||
}
|
||||
|
||||
detected = len([t for t in validated_tests if t.detection_result == TestResult.detected])
|
||||
partially = len([t for t in validated_tests if t.detection_result == TestResult.partially_detected])
|
||||
not_detected = len([t for t in validated_tests if t.detection_result == TestResult.not_detected])
|
||||
|
||||
percentage = round((detected / total) * 100, 1) if total > 0 else 0
|
||||
|
||||
return {
|
||||
"percentage": percentage,
|
||||
"detected": detected,
|
||||
"partially": partially,
|
||||
"not_detected": not_detected,
|
||||
"total": total,
|
||||
}
|
||||
|
||||
|
||||
# ── Alert Fidelity ──────────────────────────────────────────────────
|
||||
|
||||
|
||||
def calculate_alert_fidelity(db: Session) -> dict:
|
||||
"""Calculate alert fidelity: ratio of triggered detection rules."""
|
||||
total_evaluated = (
|
||||
db.query(func.count(TestDetectionResult.id))
|
||||
.filter(TestDetectionResult.triggered.isnot(None))
|
||||
.scalar()
|
||||
) or 0
|
||||
|
||||
triggered = (
|
||||
db.query(func.count(TestDetectionResult.id))
|
||||
.filter(TestDetectionResult.triggered == True)
|
||||
.scalar()
|
||||
) or 0
|
||||
|
||||
not_triggered = total_evaluated - triggered
|
||||
|
||||
return {
|
||||
"percentage": round((triggered / total_evaluated) * 100, 1) if total_evaluated > 0 else 0,
|
||||
"triggered": triggered,
|
||||
"not_triggered": not_triggered,
|
||||
"total_evaluated": total_evaluated,
|
||||
}
|
||||
|
||||
|
||||
# ── Coverage Velocity ────────────────────────────────────────────────
|
||||
|
||||
|
||||
def calculate_coverage_velocity(db: Session) -> dict:
|
||||
"""Calculate techniques validated per week."""
|
||||
# Count techniques that changed to validated/partial in the last 12 weeks
|
||||
twelve_weeks_ago = datetime.utcnow() - timedelta(weeks=12)
|
||||
|
||||
weekly_counts = (
|
||||
db.query(
|
||||
func.date_trunc("week", Technique.last_review_date).label("week"),
|
||||
func.count(Technique.id).label("count"),
|
||||
)
|
||||
.filter(
|
||||
Technique.last_review_date >= twelve_weeks_ago,
|
||||
Technique.last_review_date.isnot(None),
|
||||
)
|
||||
.group_by(func.date_trunc("week", Technique.last_review_date))
|
||||
.order_by("week")
|
||||
.all()
|
||||
)
|
||||
|
||||
if weekly_counts:
|
||||
counts = [row.count for row in weekly_counts]
|
||||
avg_per_week = round(sum(counts) / len(counts), 1)
|
||||
# Trend: compare last 4 weeks vs previous 4 weeks
|
||||
recent = counts[-4:] if len(counts) >= 4 else counts
|
||||
earlier = counts[-8:-4] if len(counts) >= 8 else counts[:len(counts) // 2] if counts else []
|
||||
|
||||
recent_avg = sum(recent) / len(recent) if recent else 0
|
||||
earlier_avg = sum(earlier) / len(earlier) if earlier else 0
|
||||
|
||||
if recent_avg > earlier_avg * 1.1:
|
||||
trend = "improving"
|
||||
elif recent_avg < earlier_avg * 0.9:
|
||||
trend = "declining"
|
||||
else:
|
||||
trend = "stable"
|
||||
else:
|
||||
avg_per_week = 0
|
||||
trend = "stable"
|
||||
|
||||
return {
|
||||
"techniques_per_week": avg_per_week,
|
||||
"trend": trend,
|
||||
}
|
||||
|
||||
|
||||
# ── Validation Throughput ────────────────────────────────────────────
|
||||
|
||||
|
||||
def calculate_validation_throughput(db: Session) -> dict:
|
||||
"""Calculate tests validated/rejected per week."""
|
||||
twelve_weeks_ago = datetime.utcnow() - timedelta(weeks=12)
|
||||
|
||||
# Tests validated
|
||||
validated_weekly = (
|
||||
db.query(
|
||||
func.date_trunc("week", Test.red_validated_at).label("week"),
|
||||
func.count(Test.id).label("count"),
|
||||
)
|
||||
.filter(
|
||||
Test.red_validated_at >= twelve_weeks_ago,
|
||||
Test.state.in_([TestState.validated, TestState.rejected]),
|
||||
)
|
||||
.group_by(func.date_trunc("week", Test.red_validated_at))
|
||||
.order_by("week")
|
||||
.all()
|
||||
)
|
||||
|
||||
if validated_weekly:
|
||||
counts = [row.count for row in validated_weekly]
|
||||
avg_per_week = round(sum(counts) / len(counts), 1)
|
||||
recent = counts[-4:] if len(counts) >= 4 else counts
|
||||
earlier = counts[-8:-4] if len(counts) >= 8 else counts[:len(counts) // 2] if counts else []
|
||||
|
||||
recent_avg = sum(recent) / len(recent) if recent else 0
|
||||
earlier_avg = sum(earlier) / len(earlier) if earlier else 0
|
||||
|
||||
if recent_avg > earlier_avg * 1.1:
|
||||
trend = "improving"
|
||||
elif recent_avg < earlier_avg * 0.9:
|
||||
trend = "declining"
|
||||
else:
|
||||
trend = "stable"
|
||||
else:
|
||||
avg_per_week = 0
|
||||
trend = "stable"
|
||||
|
||||
return {
|
||||
"tests_per_week": avg_per_week,
|
||||
"trend": trend,
|
||||
}
|
||||
|
||||
|
||||
# ── Rejection Rate ──────────────────────────────────────────────────
|
||||
|
||||
|
||||
def calculate_rejection_rate(db: Session) -> dict:
|
||||
"""Calculate rejection rate, broken down by red_lead and blue_lead."""
|
||||
validated_count = (
|
||||
db.query(func.count(Test.id))
|
||||
.filter(Test.state == TestState.validated)
|
||||
.scalar()
|
||||
) or 0
|
||||
|
||||
rejected_count = (
|
||||
db.query(func.count(Test.id))
|
||||
.filter(Test.state == TestState.rejected)
|
||||
.scalar()
|
||||
) or 0
|
||||
|
||||
total = validated_count + rejected_count
|
||||
overall_pct = round((rejected_count / total) * 100, 1) if total > 0 else 0
|
||||
|
||||
# By red_lead (red_validation_status == "rejected")
|
||||
red_rejected = (
|
||||
db.query(func.count(Test.id))
|
||||
.filter(Test.red_validation_status == "rejected")
|
||||
.scalar()
|
||||
) or 0
|
||||
red_total = (
|
||||
db.query(func.count(Test.id))
|
||||
.filter(Test.red_validation_status.in_(["approved", "rejected"]))
|
||||
.scalar()
|
||||
) or 0
|
||||
red_pct = round((red_rejected / red_total) * 100, 1) if red_total > 0 else 0
|
||||
|
||||
# By blue_lead
|
||||
blue_rejected = (
|
||||
db.query(func.count(Test.id))
|
||||
.filter(Test.blue_validation_status == "rejected")
|
||||
.scalar()
|
||||
) or 0
|
||||
blue_total = (
|
||||
db.query(func.count(Test.id))
|
||||
.filter(Test.blue_validation_status.in_(["approved", "rejected"]))
|
||||
.scalar()
|
||||
) or 0
|
||||
blue_pct = round((blue_rejected / blue_total) * 100, 1) if blue_total > 0 else 0
|
||||
|
||||
return {
|
||||
"percentage": overall_pct,
|
||||
"by_red_lead": red_pct,
|
||||
"by_blue_lead": blue_pct,
|
||||
}
|
||||
|
||||
|
||||
# ── Aggregated Operational Metrics ───────────────────────────────────
|
||||
|
||||
|
||||
def get_all_operational_metrics(db: Session) -> dict:
|
||||
"""Get all operational metrics in a single response."""
|
||||
return {
|
||||
"mttd": calculate_mttd(db),
|
||||
"mttr": calculate_mttr(db),
|
||||
"detection_efficacy": calculate_detection_efficacy(db),
|
||||
"alert_fidelity": calculate_alert_fidelity(db),
|
||||
"coverage_velocity": calculate_coverage_velocity(db),
|
||||
"validation_throughput": calculate_validation_throughput(db),
|
||||
"rejection_rate": calculate_rejection_rate(db),
|
||||
}
|
||||
|
||||
|
||||
# ── Trend Data ───────────────────────────────────────────────────────
|
||||
|
||||
|
||||
def get_operational_trend(db: Session, period: str = "90d") -> list:
|
||||
"""Get weekly trend data for operational metrics."""
|
||||
now = datetime.utcnow()
|
||||
if period == "30d":
|
||||
start = now - timedelta(days=30)
|
||||
elif period == "1y":
|
||||
start = now - timedelta(days=365)
|
||||
else:
|
||||
start = now - timedelta(days=90)
|
||||
|
||||
# Build weekly data points
|
||||
data_points = []
|
||||
current = start
|
||||
while current < now:
|
||||
week_end = min(current + timedelta(days=7), now)
|
||||
|
||||
# Detection efficacy for tests validated up to this week
|
||||
validated_up_to = (
|
||||
db.query(Test)
|
||||
.filter(
|
||||
Test.state == TestState.validated,
|
||||
Test.red_validated_at <= week_end,
|
||||
)
|
||||
.all()
|
||||
)
|
||||
|
||||
total = len(validated_up_to)
|
||||
detected = len([t for t in validated_up_to if t.detection_result == TestResult.detected])
|
||||
efficacy = round((detected / total) * 100, 1) if total > 0 else 0
|
||||
|
||||
data_points.append({
|
||||
"date": current.strftime("%Y-%m-%d"),
|
||||
"detection_efficacy": efficacy,
|
||||
"validated_tests": total,
|
||||
"detected_tests": detected,
|
||||
})
|
||||
|
||||
current = week_end
|
||||
|
||||
return data_points
|
||||
|
||||
|
||||
# ── By Team ──────────────────────────────────────────────────────────
|
||||
|
||||
|
||||
def get_metrics_by_team(db: Session) -> dict:
|
||||
"""Get metrics broken down by Red vs Blue team."""
|
||||
# Red team metrics
|
||||
red_tests_completed = (
|
||||
db.query(func.count(Test.id))
|
||||
.filter(Test.state.in_([
|
||||
TestState.blue_evaluating,
|
||||
TestState.in_review,
|
||||
TestState.validated,
|
||||
TestState.rejected,
|
||||
]))
|
||||
.scalar()
|
||||
) or 0
|
||||
|
||||
red_avg_time = None
|
||||
red_times = []
|
||||
# Time for red team to complete their phase
|
||||
tests_with_red = (
|
||||
db.query(Test)
|
||||
.filter(Test.red_validated_at.isnot(None), Test.created_at.isnot(None))
|
||||
.all()
|
||||
)
|
||||
for t in tests_with_red:
|
||||
hours = (t.red_validated_at - t.created_at).total_seconds() / 3600
|
||||
if hours > 0:
|
||||
red_times.append(hours)
|
||||
if red_times:
|
||||
red_avg_time = round(sum(red_times) / len(red_times), 1)
|
||||
|
||||
# Blue team metrics
|
||||
blue_tests_completed = (
|
||||
db.query(func.count(Test.id))
|
||||
.filter(Test.state.in_([
|
||||
TestState.in_review,
|
||||
TestState.validated,
|
||||
TestState.rejected,
|
||||
]))
|
||||
.scalar()
|
||||
) or 0
|
||||
|
||||
blue_avg_time = None
|
||||
blue_times = []
|
||||
tests_with_blue = (
|
||||
db.query(Test)
|
||||
.filter(
|
||||
Test.blue_validated_at.isnot(None),
|
||||
Test.red_validated_at.isnot(None),
|
||||
)
|
||||
.all()
|
||||
)
|
||||
for t in tests_with_blue:
|
||||
hours = (t.blue_validated_at - t.red_validated_at).total_seconds() / 3600
|
||||
if hours > 0:
|
||||
blue_times.append(hours)
|
||||
if blue_times:
|
||||
blue_avg_time = round(sum(blue_times) / len(blue_times), 1)
|
||||
|
||||
return {
|
||||
"red_team": {
|
||||
"tests_completed": red_tests_completed,
|
||||
"avg_completion_hours": red_avg_time,
|
||||
"rejection_rate": calculate_rejection_rate(db)["by_red_lead"],
|
||||
},
|
||||
"blue_team": {
|
||||
"tests_completed": blue_tests_completed,
|
||||
"avg_completion_hours": blue_avg_time,
|
||||
"rejection_rate": calculate_rejection_rate(db)["by_blue_lead"],
|
||||
},
|
||||
}
|
||||
Reference in New Issue
Block a user