Spaces:
Paused
Paused
File size: 21,747 Bytes
4a2ab42 4ae946d 4a2ab42 4ae946d 4a2ab42 4ae946d 4a2ab42 4ae946d 4a2ab42 4ae946d 4a2ab42 4ae946d 4a2ab42 4ae946d 4a2ab42 4ae946d 4a2ab42 4ae946d 4a2ab42 4ae946d 4a2ab42 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 | """
Fraud Detection Engine - Rule-Based Detection System
Task 4.1 from Orchestration Plan
This module implements pattern-based fraud detection algorithms:
- Structuring detection (transactions split to avoid reporting thresholds)
- Round-trip transactions (circular money flow)
- Velocity checks (rapid transaction patterns)
- Risk scoring system (0-100)
"""
from dataclasses import dataclass
from datetime import datetime, timedelta
from enum import Enum
from typing import Any
class FraudType(Enum):
"""Types of fraud patterns detected"""
STRUCTURING = "structuring"
ROUND_TRIP = "round_trip"
VELOCITY = "velocity"
UNUSUAL_PATTERN = "unusual_pattern"
HIGH_RISK_JURISDICTION = "high_risk_jurisdiction"
@dataclass
class Transaction:
"""Transaction data model"""
id: str
amount: float
timestamp: datetime
source_account: str
destination_account: str
description: str
merchant: str = ""
category: str = ""
metadata: dict[str, Any] = None
@dataclass
class FraudAlert:
"""Fraud detection alert"""
alert_id: str
fraud_type: FraudType
risk_score: int # 0-100
confidence: float # 0.0-1.0
transactions: list[str] # Transaction IDs
description: str
detected_at: datetime
details: dict[str, Any]
class FraudDetectionEngine:
"""
Rule-based fraud detection engine
Detects common fraud patterns in transaction data:
- Structuring: Multiple transactions just below reporting threshold
- Round-trip: Money flowing in a circle back to origin
- Velocity: Too many transactions in short time period
"""
# Default thresholds (fallback if DB fails or empty)
DEFAULT_CONFIG = {
"STRUCTURING_THRESHOLD": 10000,
"STRUCTURING_WINDOW_HOURS": 24,
"STRUCTURING_MIN_TRANSACTIONS": 3,
"VELOCITY_MAX_TRANSACTIONS": 10,
"VELOCITY_WINDOW_MINUTES": 60,
"ROUND_TRIP_MAX_HOPS": 5,
"ROUND_TRIP_TIME_WINDOW_HOURS": 72,
}
def __init__(self, db_session=None):
self.alerts: list[FraudAlert] = []
self.db = db_session
self._load_config()
def _load_config(self):
"""Load configuration from DB or use defaults"""
self.config = dict(self.DEFAULT_CONFIG)
if self.db:
try:
from core.database import FraudRule
rules = self.db.query(FraudRule).filter(FraudRule.is_active).all()
for rule in rules:
if rule.rule_id in self.config:
# Simple casting based on value_type
if rule.value_type == "int":
self.config[rule.rule_id] = int(rule.value)
elif rule.value_type == "float":
self.config[rule.rule_id] = float(rule.value)
elif rule.value_type == "json":
import json
self.config[rule.rule_id] = json.loads(rule.value)
except Exception:
# Fallback to defaults if DB fails
pass
@property
def structuring_threshold(self):
return self.config["STRUCTURING_THRESHOLD"]
@property
def structuring_window_hours(self):
return self.config["STRUCTURING_WINDOW_HOURS"]
@property
def structuring_min_transactions(self):
return self.config["STRUCTURING_MIN_TRANSACTIONS"]
@property
def velocity_max_transactions(self):
return self.config["VELOCITY_MAX_TRANSACTIONS"]
@property
def velocity_window_minutes(self):
return self.config["VELOCITY_WINDOW_MINUTES"]
@property
def round_trip_max_hops(self):
return self.config["ROUND_TRIP_MAX_HOPS"]
@property
def round_trip_time_window_hours(self):
return self.config["ROUND_TRIP_TIME_WINDOW_HOURS"]
def analyze_transactions(self, transactions: list[Transaction]) -> list[FraudAlert]:
"""
Analyze a batch of transactions for fraud patterns
Args:
transactions: List of transactions to analyze
Returns:
List of fraud alerts detected
Raises:
ValueError: If transactions is empty or None
TypeError: If transactions is not a list or contains invalid items
"""
# Input validation
if transactions is None:
raise ValueError("Transactions list cannot be None")
if not isinstance(transactions, list):
raise TypeError(f"Transactions must be a list, got {type(transactions).__name__}")
if len(transactions) == 0:
raise ValueError("Transactions list cannot be empty")
# Validate each transaction
for i, tx in enumerate(transactions):
if not isinstance(tx, Transaction):
raise TypeError(f"Transaction at index {i} must be a Transaction instance, got {type(tx).__name__}")
self.alerts = []
# Run all detection algorithms
self.alerts.extend(self._detect_structuring(transactions))
self.alerts.extend(self._detect_velocity(transactions))
self.alerts.extend(self._detect_round_trips(transactions))
return self.alerts
def _detect_structuring(self, transactions: list[Transaction]) -> list[FraudAlert]:
"""
Detect structuring: Multiple transactions just below reporting threshold
"""
account_groups = self._group_transactions_by_account_pairs(transactions)
return self._analyze_account_groups_for_structuring(account_groups)
def _group_transactions_by_account_pairs(
self, transactions: list[Transaction]
) -> dict[tuple[str, str], list[Transaction]]:
"""Group transactions by account pairs for structuring analysis"""
account_groups: dict[tuple[str, str], list[Transaction]] = {}
for tx in transactions:
account_pair = tuple(sorted([tx.source_account, tx.destination_account]))
if account_pair not in account_groups:
account_groups[account_pair] = []
account_groups[account_pair].append(tx)
return account_groups
def _analyze_account_groups_for_structuring(
self, account_groups: dict[tuple[str, str], list[Transaction]]
) -> list[FraudAlert]:
"""Analyze grouped transactions for structuring patterns"""
alerts = []
for account_pair, txs in account_groups.items():
txs.sort(key=lambda x: x.timestamp)
structuring_windows = self._find_structuring_windows(txs)
for window_data in structuring_windows:
if self._is_significant_structuring(window_data):
alert = self._create_structuring_alert(account_pair, window_data)
alerts.append(alert)
return alerts
def _find_structuring_windows(self, transactions: list[Transaction]) -> list[dict[str, Any]]:
"""Find windows of transactions that may indicate structuring"""
windows = []
for i in range(len(transactions)):
window_txs = []
window_start = transactions[i].timestamp
total_amount = 0
# Collect transactions within time window
for tx in transactions[i:]:
if self._is_within_structuring_window(tx, window_start):
if self._is_suspicious_amount(tx.amount):
window_txs.append(tx)
total_amount += tx.amount
else:
break
if window_txs:
windows.append(
{
"transactions": window_txs,
"total_amount": total_amount,
"start_time": window_start,
"end_time": window_txs[-1].timestamp,
}
)
return windows
def _is_within_structuring_window(self, transaction: Transaction, window_start) -> bool:
"""Check if transaction is within structuring time window"""
return (transaction.timestamp - window_start).total_seconds() / 3600 <= self.STRUCTURING_WINDOW_HOURS
def _is_suspicious_amount(self, amount: float) -> bool:
"""Check if transaction amount is suspiciously close to threshold"""
return 0.8 * self.STRUCTURING_THRESHOLD <= amount < self.STRUCTURING_THRESHOLD
def _is_significant_structuring(self, window_data: dict[str, Any]) -> bool:
"""Determine if a transaction window represents significant structuring"""
txs = window_data["transactions"]
total_amount = window_data["total_amount"]
return len(txs) >= self.STRUCTURING_MIN_TRANSACTIONS and total_amount >= self.STRUCTURING_THRESHOLD
def _create_structuring_alert(self, account_pair: tuple[str, str], window_data: dict[str, Any]) -> FraudAlert:
"""Create a structuring fraud alert"""
txs = window_data["transactions"]
total_amount = window_data["total_amount"]
# Calculate risk score based on various factors
base_score = 70 # Structuring is inherently suspicious
proximity_bonus = self._calculate_proximity_bonus(txs)
count_bonus = min(len(txs) * 5, 30) # Up to 30 points for many transactions
risk_score = min(base_score + proximity_bonus + count_bonus, 100)
return FraudAlert(
alert_id=f"STRUCTURING_{account_pair[0]}_{account_pair[1]}_{int(window_data['start_time'].timestamp())}",
case_id=None, # Will be assigned by caller
title="Potential Structuring Detected",
description=self._generate_structuring_description(txs, total_amount),
severity=self._calculate_severity(risk_score),
risk_score=risk_score,
alert_type="structuring",
entities=[
{"type": "account", "value": account_pair[0], "confidence": 0.9},
{"type": "account", "value": account_pair[1], "confidence": 0.9},
],
metadata={
"transaction_count": len(txs),
"total_amount": total_amount,
"time_window_hours": self.STRUCTURING_WINDOW_HOURS,
"threshold": self.STRUCTURING_THRESHOLD,
},
confidence=0.85,
detection_method="transaction_pattern_analysis",
)
def _calculate_proximity_bonus(self, transactions: list[Transaction]) -> int:
"""Calculate bonus based on how close amounts are to threshold"""
if not transactions:
return 0
# Average proximity to threshold (0-40 points)
proximities = []
for tx in transactions:
if tx.amount < self.STRUCTURING_THRESHOLD:
proximity = (self.STRUCTURING_THRESHOLD - tx.amount) / self.STRUCTURING_THRESHOLD
proximities.append(proximity)
avg_proximity = sum(proximities) / len(proximities) if proximities else 0
return int(avg_proximity * 40)
def _generate_structuring_description(self, transactions: list[Transaction], total_amount: float) -> str:
"""Generate human-readable description of structuring pattern"""
count = len(transactions)
avg_amount = total_amount / count
return (
f"Detected {count} transactions totaling ${total_amount:,.2f} "
f"(avg: ${avg_amount:,.2f}) within {self.STRUCTURING_WINDOW_HOURS} hours. "
f"All amounts suspiciously close to ${self.STRUCTURING_THRESHOLD:,.0f} CTR threshold."
)
def _calculate_severity(self, risk_score: int) -> str:
"""Convert risk score to severity level"""
if risk_score >= 90:
return "critical"
elif risk_score >= 75:
return "high"
elif risk_score >= 60:
return "medium"
else:
return "low"
return alerts
def _detect_velocity(self, transactions: list[Transaction]) -> list[FraudAlert]:
"""
Detect velocity fraud: Too many transactions in short time period
Pattern: Unusual burst of transaction activity that may indicate:
- Account takeover
- Card testing
- Automated fraud
"""
alerts = []
# Group by source account
account_txs: dict[str, list[Transaction]] = {}
for tx in transactions:
if tx.source_account not in account_txs:
account_txs[tx.source_account] = []
account_txs[tx.source_account].append(tx)
# Check each account for velocity issues
for account, txs in account_txs.items():
txs.sort(key=lambda x: x.timestamp)
# Sliding window to detect bursts
for i in range(len(txs)):
window_start = txs[i].timestamp
window_txs = []
for tx in txs[i:]:
if (tx.timestamp - window_start).total_seconds() / 60 <= self.VELOCITY_WINDOW_MINUTES:
window_txs.append(tx)
else:
break
# Alert if too many transactions in window
if len(window_txs) >= self.VELOCITY_MAX_TRANSACTIONS:
# Calculate risk score based on velocity
base_score = 50
velocity_bonus = min((len(window_txs) - self.VELOCITY_MAX_TRANSACTIONS) * 5, 40)
# Bonus for very short time window
actual_window_minutes = (window_txs[-1].timestamp - window_txs[0].timestamp).total_seconds() / 60
time_bonus = int((1 - actual_window_minutes / self.VELOCITY_WINDOW_MINUTES) * 10)
risk_score = min(base_score + velocity_bonus + time_bonus, 100)
alert = FraudAlert(
alert_id=f"VEL_{account}_{window_txs[0].id}",
fraud_type=FraudType.VELOCITY,
risk_score=risk_score,
confidence=0.75,
transactions=[tx.id for tx in window_txs],
description=f"High velocity detected: {len(window_txs)} transactions from account {account} "
f"in {actual_window_minutes:.1f} minutes",
detected_at=datetime.now(),
details={
"account": account,
"transaction_count": len(window_txs),
"window_minutes": actual_window_minutes,
"threshold": self.VELOCITY_MAX_TRANSACTIONS,
},
)
alerts.append(alert)
break # One alert per account per window
return alerts
def _detect_round_trips(self, transactions: list[Transaction]) -> list[FraudAlert]:
"""
Detect round-trip transactions: Money flowing in a circle
Pattern: A β B β C β A (money returns to origin)
Often used for:
- Money laundering (layering phase)
- Creating false business activity
- Tax evasion
"""
alerts = []
# Build transaction graph
graph: dict[str, list[tuple[str, Transaction]]] = {}
for tx in transactions:
if tx.source_account not in graph:
graph[tx.source_account] = []
graph[tx.source_account].append((tx.destination_account, tx))
# Find cycles using DFS
def find_cycle(start: str, current: str, path: list[Transaction], visited: set) -> list[Transaction]:
"""Find cycle from current node back to start"""
if current == start and len(path) > 1:
return path # Found cycle!
if current in visited or len(path) > self.ROUND_TRIP_MAX_HOPS:
return []
visited.add(current)
if current in graph:
for next_account, tx in graph[current]:
# Check if within time window
if (
path
and (tx.timestamp - path[0].timestamp).total_seconds() / 3600
> self.ROUND_TRIP_TIME_WINDOW_HOURS
):
continue
result = find_cycle(start, next_account, [*path, tx], visited.copy())
if result:
return result
return []
# Check each account for round trips
checked_cycles = set()
for start_account in graph:
cycle = find_cycle(start_account, start_account, [], set())
if cycle:
# Create unique cycle identifier
cycle_id = "_".join(sorted([tx.id for tx in cycle]))
if cycle_id not in checked_cycles:
checked_cycles.add(cycle_id)
total_amount = sum(tx.amount for tx in cycle)
# Calculate risk score
base_score = 70 # Round trips are inherently suspicious
hop_bonus = len(cycle) * 5 # More hops = more suspicious
amount_bonus = min(int(total_amount / 10000) * 5, 20)
risk_score = min(base_score + hop_bonus + amount_bonus, 100)
# Build path description
path_desc = " β ".join([cycle[0].source_account] + [tx.destination_account for tx in cycle])
alert = FraudAlert(
alert_id=f"ROUND_{cycle_id}",
fraud_type=FraudType.ROUND_TRIP,
risk_score=risk_score,
confidence=0.90,
transactions=[tx.id for tx in cycle],
description=f"Round-trip detected: {len(cycle)} transactions forming cycle {path_desc}, "
f"total amount: ${total_amount:,.2f}",
detected_at=datetime.now(),
details={
"cycle_length": len(cycle),
"total_amount": total_amount,
"path": path_desc,
"time_span_hours": (cycle[-1].timestamp - cycle[0].timestamp).total_seconds() / 3600,
},
)
alerts.append(alert)
return alerts
def calculate_overall_risk(self, account: str, transactions: list[Transaction]) -> int:
"""
Calculate overall risk score for an account (0-100)
Considers:
- Number of fraud alerts
- Severity of alerts
- Transaction patterns
"""
account_txs = [tx for tx in transactions if tx.source_account == account or tx.destination_account == account]
if not account_txs:
return 0
# Analyze account's transactions
alerts = self.analyze_transactions(account_txs)
account_alerts = [a for a in alerts if account in str(a.details)]
if not account_alerts:
return 10 # Base risk for any active account
# Weighted average of alert scores
total_score = sum(alert.risk_score * alert.confidence for alert in account_alerts)
avg_score = int(total_score / len(account_alerts))
# Bonus for multiple different fraud types
fraud_types = len({a.fraud_type for a in account_alerts})
diversity_bonus = min(fraud_types * 10, 30)
return min(avg_score + diversity_bonus, 100)
# Example usage and testing
if __name__ == "__main__":
# Create test transactions
test_txs = [
# Structuring example: 3 transactions just below $10k threshold
Transaction(
"tx1",
9900,
datetime.now() - timedelta(hours=2),
"ACC001",
"ACC002",
"Payment 1",
),
Transaction(
"tx2",
9800,
datetime.now() - timedelta(hours=1),
"ACC001",
"ACC002",
"Payment 2",
),
Transaction("tx3", 9700, datetime.now(), "ACC001", "ACC002", "Payment 3"),
# Velocity example: 12 transactions in 30 minutes
*[
Transaction(
f"vtx{i}",
100,
datetime.now() - timedelta(minutes=30 - i * 2),
"ACC003",
f"SHOP{i}",
f"Purchase {i}",
)
for i in range(12)
],
# Round trip example: A β B β C β A
Transaction(
"rtx1",
5000,
datetime.now() - timedelta(hours=48),
"ACC004",
"ACC005",
"Transfer",
),
Transaction(
"rtx2",
4800,
datetime.now() - timedelta(hours=24),
"ACC005",
"ACC006",
"Payment",
),
Transaction("rtx3", 4600, datetime.now(), "ACC006", "ACC004", "Return"),
]
# Run detection
engine = FraudDetectionEngine()
alerts = engine.analyze_transactions(test_txs)
print(f"\nπ¨ Detected {len(alerts)} fraud alerts:\n")
for alert in alerts:
print(f"Alert ID: {alert.alert_id}")
print(f"Type: {alert.fraud_type.value}")
print(f"Risk Score: {alert.risk_score}/100")
print(f"Confidence: {alert.confidence:.0%}")
print(f"Description: {alert.description}")
print(f"Transactions: {len(alert.transactions)}")
print("-" * 80)
|