minhan6559's picture
Upload 102 files
9e3d618 verified
#!/usr/bin/env python3
"""
Evaluate system performance metrics.
Calculates detection rates, coverage, accuracy, and overall effectiveness
based on tactic occurrence counts. Generates separate reports for each model.
Usage:
python evaluate_metrics.py [--input INPUT_PATH] [--output OUTPUT_PATH]
"""
import argparse
import json
from pathlib import Path
from typing import Dict, List, Any
from datetime import datetime
import statistics
class SystemEvaluator:
"""Evaluates multi-agent system performance"""
def __init__(self, tactic_counts_file: Path):
self.tactic_counts_file = tactic_counts_file
self.tactic_data = []
self.load_tactic_counts()
def load_tactic_counts(self):
"""Load tactic counts summary data"""
if not self.tactic_counts_file.exists():
raise FileNotFoundError(
f"Tactic counts file not found: {self.tactic_counts_file}"
)
data = json.loads(self.tactic_counts_file.read_text(encoding="utf-8"))
self.tactic_data = data.get("results", [])
print(f"[INFO] Loaded {len(self.tactic_data)} tactic analysis results")
def group_by_model(self) -> Dict[str, List[Dict]]:
"""Group tactic data by model"""
models = {}
for item in self.tactic_data:
model = item["model"]
if model not in models:
models[model] = []
models[model].append(item)
return models
def calculate_detection_rate(self, model_data: List[Dict] = None) -> Dict[str, Any]:
"""Calculate detection rate: % of files where tactic was correctly detected"""
data_to_use = model_data if model_data is not None else self.tactic_data
# Aggregate by tactic
tactic_aggregates = {}
for item in data_to_use:
tactic = item["tactic"]
if tactic not in tactic_aggregates:
tactic_aggregates[tactic] = {
"total_files": 0,
"files_detected": 0,
"total_events": 0,
}
tactic_aggregates[tactic]["total_files"] += 1
tactic_aggregates[tactic]["files_detected"] += item["tactic_detected"]
tactic_aggregates[tactic]["total_events"] += item[
"total_abnormal_events_detected"
]
total_files = sum(agg["total_files"] for agg in tactic_aggregates.values())
total_detected = sum(
agg["files_detected"] for agg in tactic_aggregates.values()
)
total_events = sum(agg["total_events"] for agg in tactic_aggregates.values())
per_tactic_detection = []
for tactic, agg in sorted(tactic_aggregates.items()):
files = agg["total_files"]
detected = agg["files_detected"]
events = agg["total_events"]
detection_rate = (detected / files * 100) if files > 0 else 0.0
per_tactic_detection.append(
{
"tactic": tactic,
"total_files": files,
"files_detected": detected,
"files_missed": files - detected,
"total_abnormal_events_detected": events,
"detection_rate_percent": detection_rate,
"status": (
"GOOD"
if detection_rate >= 50
else ("POOR" if detection_rate > 0 else "NONE")
),
}
)
overall_detection_rate = (
(total_detected / total_files * 100) if total_files > 0 else 0.0
)
return {
"overall_detection_rate_percent": overall_detection_rate,
"total_files": total_files,
"total_files_detected": total_detected,
"total_files_missed": total_files - total_detected,
"total_abnormal_events_detected": total_events,
"total_tactics": len(tactic_aggregates),
"per_tactic_detection": per_tactic_detection,
}
def calculate_coverage(self, model_data: List[Dict] = None) -> Dict[str, Any]:
"""Calculate coverage: how many tactics have at least one successful detection"""
data_to_use = model_data if model_data is not None else self.tactic_data
# Aggregate by tactic
tactic_aggregates = {}
for item in data_to_use:
tactic = item["tactic"]
if tactic not in tactic_aggregates:
tactic_aggregates[tactic] = 0
tactic_aggregates[tactic] += item["tactic_detected"]
total_tactics = len(tactic_aggregates)
tactics_with_detection = sum(
1 for count in tactic_aggregates.values() if count > 0
)
tactics_with_zero_detection = total_tactics - tactics_with_detection
coverage_percent = (
(tactics_with_detection / total_tactics * 100) if total_tactics > 0 else 0.0
)
detected_tactics = sorted(
[tactic for tactic, count in tactic_aggregates.items() if count > 0]
)
missed_tactics = sorted(
[tactic for tactic, count in tactic_aggregates.items() if count == 0]
)
return {
"coverage_percent": coverage_percent,
"total_tactics_tested": total_tactics,
"tactics_with_detection": tactics_with_detection,
"tactics_with_zero_detection": tactics_with_zero_detection,
"detected_tactics": detected_tactics,
"missed_tactics": missed_tactics,
}
def calculate_accuracy_proxy(self, model_data: List[Dict] = None) -> Dict[str, Any]:
"""Calculate accuracy proxy: detection success rate per tactic"""
data_to_use = model_data if model_data is not None else self.tactic_data
# Aggregate by tactic
tactic_aggregates = {}
for item in data_to_use:
tactic = item["tactic"]
if tactic not in tactic_aggregates:
tactic_aggregates[tactic] = {"total_files": 0, "files_detected": 0}
tactic_aggregates[tactic]["total_files"] += 1
tactic_aggregates[tactic]["files_detected"] += item["tactic_detected"]
accuracy_scores = []
for tactic, agg in sorted(tactic_aggregates.items()):
if agg["total_files"] > 0:
accuracy = agg["files_detected"] / agg["total_files"]
accuracy_scores.append(
{
"tactic": tactic,
"accuracy_score": accuracy,
"interpretation": (
"Perfect"
if accuracy == 1.0
else ("Partial" if accuracy > 0 else "Failed")
),
}
)
avg_accuracy = (
statistics.mean([s["accuracy_score"] for s in accuracy_scores])
if accuracy_scores
else 0.0
)
return {
"average_accuracy_score": avg_accuracy,
"per_tactic_accuracy": accuracy_scores,
"perfect_matches": sum(
1 for s in accuracy_scores if s["accuracy_score"] == 1.0
),
"partial_matches": sum(
1 for s in accuracy_scores if 0 < s["accuracy_score"] < 1.0
),
"failed_matches": sum(
1 for s in accuracy_scores if s["accuracy_score"] == 0.0
),
}
def calculate_effectiveness(self, model_data: List[Dict] = None) -> Dict[str, Any]:
"""Calculate overall system effectiveness score (0-100)"""
detection = self.calculate_detection_rate(model_data)
coverage = self.calculate_coverage(model_data)
accuracy = self.calculate_accuracy_proxy(model_data)
# Weighted effectiveness score
# 40% detection rate, 30% coverage, 30% accuracy
effectiveness_score = (
detection["overall_detection_rate_percent"] * 0.4
+ coverage["coverage_percent"] * 0.3
+ accuracy["average_accuracy_score"] * 100 * 0.3
)
# Grade the system
if effectiveness_score >= 80:
grade = "EXCELLENT"
elif effectiveness_score >= 60:
grade = "GOOD"
elif effectiveness_score >= 40:
grade = "FAIR"
elif effectiveness_score >= 20:
grade = "POOR"
else:
grade = "CRITICAL"
return {
"effectiveness_score": effectiveness_score,
"grade": grade,
"component_scores": {
"detection_rate": detection["overall_detection_rate_percent"],
"coverage_rate": coverage["coverage_percent"],
"accuracy_score": accuracy["average_accuracy_score"] * 100,
},
}
def identify_issues(self, model_data: List[Dict] = None) -> List[str]:
"""Identify specific issues and gaps"""
issues = []
detection = self.calculate_detection_rate(model_data)
coverage = self.calculate_coverage(model_data)
# Check overall detection
if detection["overall_detection_rate_percent"] < 20:
issues.append(
f"CRITICAL: Overall detection rate is only {detection['overall_detection_rate_percent']:.1f}%. "
f"System is failing to detect most attacks ({detection['total_files_missed']}/{detection['total_files']} files missed)."
)
elif detection["overall_detection_rate_percent"] < 50:
issues.append(
f"WARNING: Detection rate is {detection['overall_detection_rate_percent']:.1f}%, "
f"below acceptable threshold of 50% ({detection['total_files_missed']}/{detection['total_files']} files missed)."
)
# Check coverage
if coverage["tactics_with_zero_detection"] > 0:
missed = ", ".join(coverage["missed_tactics"])
issues.append(
f"COVERAGE GAP: {coverage['tactics_with_zero_detection']} tactics have zero detection: {missed}"
)
# Check for specific problematic tactics
for item in detection["per_tactic_detection"]:
if item["total_files"] > 0 and item["detection_rate_percent"] == 0:
issues.append(
f"TACTIC FAILURE: '{item['tactic']}' - "
f"{item['total_files']} files analyzed, 0 detected"
)
# Check for data quality issues
data_to_use = model_data if model_data is not None else self.tactic_data
zero_event_tactics = [
item["tactic"]
for item in data_to_use
if item["total_abnormal_events_detected"] == 0
]
if zero_event_tactics:
unique_zero = list(set(zero_event_tactics))
issues.append(
f"DATA ISSUE: No events to analyze for tactics: {', '.join(unique_zero)}"
)
if not issues:
issues.append(
"No critical issues detected. System is performing within acceptable parameters."
)
return issues
def run_evaluation_for_model(
self, model_name: str, model_data: List[Dict]
) -> Dict[str, Any]:
"""Run full evaluation for a specific model"""
print(f"\nEvaluating model: {model_name} ({len(model_data)} files)")
detection = self.calculate_detection_rate(model_data)
coverage = self.calculate_coverage(model_data)
accuracy = self.calculate_accuracy_proxy(model_data)
effectiveness = self.calculate_effectiveness(model_data)
issues = self.identify_issues(model_data)
report = {
"timestamp": datetime.now().isoformat(),
"model_name": model_name,
"evaluation_metrics": {
"detection_rate": detection,
"coverage": coverage,
"accuracy_proxy": accuracy,
"effectiveness": effectiveness,
},
"issues_identified": issues,
}
return report
def run_evaluation(self) -> Dict[str, Any]:
"""Run full evaluation and compile report for all models"""
print("\n" + "=" * 80)
print("RUNNING SYSTEM EVALUATION")
print("=" * 80 + "\n")
# Group data by model
models_data = self.group_by_model()
if not models_data:
print("[WARNING] No model data found")
return {"error": "No model data found"}
print(f"Found {len(models_data)} models: {', '.join(models_data.keys())}")
# Generate reports for each model
model_reports = {}
for model_name, model_data in models_data.items():
print(f"\nProcessing model: {model_name}")
model_reports[model_name] = self.run_evaluation_for_model(
model_name, model_data
)
# Create summary report
summary_report = {
"timestamp": datetime.now().isoformat(),
"total_models_evaluated": len(model_reports),
"models": list(model_reports.keys()),
"model_reports": model_reports,
}
return summary_report
def main():
parser = argparse.ArgumentParser(
description="Evaluate multi-agent system performance"
)
parser.add_argument(
"--input",
default="evaluation/full_pipeline/results/tactic_counts_summary.json",
help="Path to tactic_counts_summary.json",
)
parser.add_argument(
"--output",
default="evaluation/full_pipeline/results/evaluation_report.json",
help="Output file for evaluation report",
)
args = parser.parse_args()
input_path = Path(args.input)
output_path = Path(args.output)
if not input_path.exists():
print(f"[ERROR] Input file not found: {input_path}")
print("Run count_tactics.py first to generate tactic counts")
return 1
# Run evaluation
evaluator = SystemEvaluator(input_path)
report = evaluator.run_evaluation()
if "error" in report:
print(f"[ERROR] {report['error']}")
return 1
# Save main report
output_path.parent.mkdir(parents=True, exist_ok=True)
output_path.write_text(json.dumps(report, indent=2), encoding="utf-8")
# Save individual model reports
for model_name, model_report in report["model_reports"].items():
model_output_path = (
output_path.parent
/ f"evaluation_report_{model_name.replace(':', '_').replace('/', '_')}.json"
)
model_output_path.write_text(
json.dumps(model_report, indent=2), encoding="utf-8"
)
print(f"Model report saved: {model_output_path}")
# Display summary
print("\n" + "=" * 80)
print("EVALUATION COMPLETE")
print("=" * 80)
print(f"Models evaluated: {report['total_models_evaluated']}")
print(f"Models: {', '.join(report['models'])}")
# Show summary for each model
for model_name, model_report in report["model_reports"].items():
effectiveness = model_report["evaluation_metrics"]["effectiveness"]
print(f"\n{model_name}:")
print(f" Effectiveness Score: {effectiveness['effectiveness_score']:.1f}/100")
print(f" Grade: {effectiveness['grade']}")
print(
f" Detection Rate: {effectiveness['component_scores']['detection_rate']:.1f}%"
)
print(f" Coverage: {effectiveness['component_scores']['coverage_rate']:.1f}%")
print(f" Accuracy: {effectiveness['component_scores']['accuracy_score']:.1f}%")
print(f"\nMain report saved to: {output_path}")
print("=" * 80 + "\n")
return 0
if __name__ == "__main__":
exit(main())