Fraud Scoring System for Orders in 1C-Bitrix: Setup and Implementation

Every day, dozens of orders pass through an online store's cart. Some of them are fraud: criminals use stolen cards, create fake orders, test payment gateways. We have seen stores where fraud losses reached 5% of turnover. Standard Bitrix tools do not provide a flexible scoring system — we wrote our own.

We have seen stores where, due to lack of protection, fraudsters placed orders worth hundreds of about $9–13 in savings — and the store lost both goods and money. Our approach is a scoring system that evaluates each order based on a set of signals and automatically decides: allow, send for manual review, or block. Unlike primitive IP limiting, scoring considers signal combinations — this gives 5 times fewer false positives.

Analyzed Fraud Signals

Signals are divided into five categories: IP address, email, phone, order amount, and name. Each signal gives a certain number of risk points. For example:

Total score is the sum of all signals, capped at 100. Decision: allow (0–39), review (40–69), block (70–100).

How the Scoring System Works

Here is the key class that performs the check. We use raw SQL for performance:

namespace Local\Fraud;

class FraudScorer
{
    // Thresholds
    private const BLOCK_SCORE = 70;
    private const REVIEW_SCORE = 40;

    public function score(\Bitrix\Sale\Order $order): ScoreResult
    {
        $signals = [];

        $ip = $_SERVER['REMOTE_ADDR'] ?? '';
        $props = $order->getPropertyCollection();
        $email = $props->getItemByOrderPropertyCode('EMAIL')?->getValue() ?? '';
        $phone = $props->getItemByOrderPropertyCode('PHONE')?->getValue() ?? '';
        $name = trim(
            ($props->getItemByOrderPropertyCode('NAME')?->getValue() ?? '') . ' ' .
            ($props->getItemByOrderPropertyCode('LAST_NAME')?->getValue() ?? '')
        );

        // IP signals
        $signals['ip_orders_1h'] = $this->ipOrders($ip, 1) * 8; // max ~80 at 10 orders
        $signals['ip_orders_24h'] = $this->ipOrders($ip, 24) * 2; // max ~40 at 20 orders
        $signals['ip_in_stoplist'] = $this->isInStopList($ip) ? 80 : 0;

        // Email signals
        $signals['disposable_email'] = $this->isDisposableEmail($email) ? 35 : 0;
        $signals['no_email'] = empty($email) ? 25 : 0;
        $signals['email_orders_24h'] = $this->emailOrders($email, 24) * 5;

        // Phone signals
        $signals['invalid_phone'] = !$this->isValidPhone($phone) ? 20 : 0;

        // Amount and history
        $signals['high_amount_new'] = $this->highAmountNewUser($order) ? 30 : 0;
        $signals['unusual_amount'] = $this->isUnusualAmount($order, (int)$order->getUserId()) ? 15 : 0;

        // Name
        $signals['suspicious_name'] = $this->isSuspiciousName($name) ? 20 : 0;

        $total = min(100, array_sum($signals));

        return new ScoreResult(
            score: $total,
            signals: array_filter($signals),
            action: match(true) {
                $total >= self::BLOCK_SCORE => 'block',
                $total >= self::REVIEW_SCORE => 'review',
                default => 'allow',
            }
        );
    }

    private function ipOrders(string $ip, int $hours): int
    {
        $safe = \Bitrix\Main\Application::getConnection()->getSqlHelper()->forSql($ip);
        return (int)\Bitrix\Main\Application::getConnection()->query(
            "SELECT COUNT(*) cnt FROM b_sale_order WHERE CREATED_BY_IP = '{$safe}' AND DATE_INSERT > DATE_SUB(NOW(), INTERVAL {$hours} HOUR)"
        )->fetch()['cnt'];
    }

    private function isInStopList(string $ip): bool
    {
        $safe = \Bitrix\Main\Application::getConnection()->getSqlHelper()->forSql($ip);
        return (bool)\Bitrix\Main\Application::getConnection()->query(
            "SELECT ID FROM b_stop_list WHERE IP_ADDR = '{$safe}' AND ACTIVE = 'Y' LIMIT 1"
        )->fetch();
    }

    private function emailOrders(string $email, int $hours): int
    {
        if (empty($email)) return 0;
        $safe = \Bitrix\Main\Application::getConnection()->getSqlHelper()->forSql($email);
        return (int)\Bitrix\Main\Application::getConnection()->query(
            "SELECT COUNT(*) cnt FROM b_sale_order_props_value pv JOIN b_sale_order_props p ON p.ID = pv.ORDER_PROPS_ID JOIN b_sale_order o ON o.ID = pv.ORDER_ID WHERE p.CODE = 'EMAIL' AND pv.VALUE = '{$safe}' AND o.DATE_INSERT > DATE_SUB(NOW(), INTERVAL {$hours} HOUR)"
        )->fetch()['cnt'];
    }

    private function isDisposableEmail(string $email): bool
    {
        $domain = strtolower(substr(strrchr($email, '@'), 1));
        return in_array($domain, [
            'mailinator.com',
            'guerrillamail.com',
            'tempmail.com',
            'throwam.com',
            'yopmail.com',
            '10minutemail.com',
        ], true);
    }

    private function isValidPhone(string $phone): bool
    {
        $digits = preg_replace('/\D/', '', $phone);
        return strlen($digits) >= 10 && strlen($digits) <= 15;
    }

    private function highAmountNewUser(\Bitrix\Sale\Order $order): bool
    {
        $userId = (int)$order->getUserId();
        if ($order->getPrice() < 30000 || $userId <= 0) return false;
        $prevCount = (int)\Bitrix\Main\Application::getConnection()->query(
            "SELECT COUNT(*) cnt FROM b_sale_order WHERE USER_ID = {$userId}"
        )->fetch()['cnt'];
        return $prevCount === 0;
    }

    private function isUnusualAmount(\Bitrix\Sale\Order $order, int $userId): bool
    {
        if ($userId <= 0) return false;
        $avg = (float)\Bitrix\Main\Application::getConnection()->query(
            "SELECT AVG(PRICE) avg FROM b_sale_order WHERE USER_ID = {$userId} AND STATUS_ID NOT IN ('C')"
        )->fetch()['avg'];
        return $avg > 0 && $order->getPrice() > $avg * 5;
    }

    private function isSuspiciousName(string $name): bool
    {
        // Fully numeric name, too short, only special characters
        return preg_match('/^\d+$/', $name) || mb_strlen($name) < 3 || preg_match('/[<>{}\]/', $name);
    }
}

Check Result

namespace Local\Fraud;

class ScoreResult
{
    public function __construct(
        public readonly int $score,
        public readonly array $signals,
        public readonly string $action, // 'allow', 'review', 'block'
    ) {}

    public function isBlocked(): bool
    {
        return $this->action === 'block';
    }

    public function needsReview(): bool
    {
        return $this->action === 'review';
    }

    public function getComment(): string
    {
        $parts = ["[FRAUD_SCORE:{$this->score}]"];
        foreach ($this->signals as $signal => $value) {
            $parts[] = "{$signal}:{$value}";
        }
        return implode(' ', $parts);
    }
}

Logging Check Results

All checks are logged into the HL-block FraudLog for analysis and threshold tuning:

Log analysis over 2–4 weeks allows calibrating thresholds for a specific store. We will help select optimal values based on your statistics.

What's Included

When setting up fraud scoring, we:

• Deploy the code on your 1C-Bitrix project (PHP 8.1+)
• Configure the OnSaleOrderBeforeSaved event handler to call scoring before order saving
• Create the HL-block FraudLog and an admin page for viewing logs
• Integrate IP stop-list using the standard b_stop_list table
• Calibrate thresholds on historical data (at least 1 month of orders)
• Provide documentation on architecture and how to add new signals
• Offer 2 weeks of post-deployment support

Work Process and Timeline

1. Analytics — study your order scheme, identify typical fraud patterns (1 day)
2. Design — determine signal set and thresholds for your budget (1 day)
3. Implementation — write scoring system code and integration (2–3 days)
4. Testing — check on historical data and live orders (1–2 days)
5. Deploy and calibration — launch in production, refine thresholds based on actual data (up to 1 week)

Why Scoring Is 5x More Accurate Than Primitive Blocks?

Because it considers signal combinations, not single indicators. For example, a new user with a high order amount is not always fraud, but if the email is disposable and IP is in the stop-list, risk is high. This approach reduces false positives by 5 times compared to blocking on a single indicator.

How to Start Using Scoring in Your Store?

We implemented such a system in an electronics store with a turnover of $135k–195k per month. Result: 98% of fraud orders are blocked automatically, another 1.5% go to manual review. False positives are less than 0.3%. Fraud losses decreased by 70% in the first month.

Want the same? Order an audit of your current orders for fraud — we will offer a turnkey solution. Get a consultation: we will evaluate your project and set up scoring with calibration based on your statistics.

\Bitrix\Main\EventManager::getInstance()->addEventHandler(
    'sale',
    'OnSaleOrderBeforeSaved',
    function(\Bitrix\Main\Event $event) {
        $order = $event->getParameter('ENTITY');
        if (!$order instanceof \Bitrix\Sale\Order) return;
        $scorer = new \Local\Fraud\FraudScorer();
        $result = $scorer->score($order);
        if ($result->isBlocked()) {
            $order->setField('STATUS_ID', 'N');
            // do not reserve
            // optionally add comment
        }
    }
);