Account takeover indicator rule
An account takeover indicator rule detects the combination of signals that together suggest someone other than the owner is in control of an account. No single indicator is conclusive, so the rule aggregates several together, compares each against the account's baseline and scores the result: a new device, a credential or contact change, an unusual location, failed logins, and a first payment to a new payee.
The fraud scenario this rule addresses
An attacker obtains credentials through phishing, a data breach, malware or a social-engineering call. They log in, often change contact details or reset credentials so the real owner cannot take control back or receive alerts, enrol their own device, add a payee and move money out.
Each step may be explainable alone. Together, and in that order and with that timing, they describe a takeover.
Data the rule requires
- Login events with outcome, device, location and network
- Credential, contact-detail and limit change events with timestamps
- Device enrolment events
- Beneficiary additions and payments
- The account's baseline for each of these signals
Example logic
Illustrative pseudocode: it compares each event with the rule's parameters and flags the ones that match. Parameters are symbolic: every value is institution-specific and has to be derived from your own transaction distribution.
RULE account_takeover_indicators
FOR EACH account A, evaluated on each relevant event:
signals = []
IF new_device(A, within = :window): signals += "new_device"
IF credentials_or_contact_changed(A, :window): signals += "profile_change"
IF location_out_of_pattern(A, :window): signals += "location"
IF failed_logins(A, :window) >= :max_failures: signals += "failed_logins"
IF new_beneficiary_paid(A, :window): signals += "new_payee"
weight = sum(:weights[s] FOR s IN signals)
IF weight >= :alert_weight
AND count(signals) >= :min_signals:
RAISE alert(rule = "account_takeover_indicators",
account = A, evidence = signals + events)
Choosing a threshold
This is a composite rule, so the trade-off is between sensitivity and the number of independent signals required. Requiring many signals is precise but waits until the attacker has progressed far, perhaps until after the money moved. Requiring only one or two alerts early but resembles a rule on any of them alone.
Weights express how much each signal matters. A change of contact details followed by a payment is worth far more than a new device alone, and weights are a judgement that needs revisiting as attackers change method. The window is a second lever: too short and the attacker spaces out the steps, too long and ordinary life accumulates signals into false alerts.
An ordering requirement, such as the profile change preceding the payment, makes the rule sharper than a bare count, and also makes it easier to evade, so whether to require order is a design decision.
False positives to expect
A genuine customer can produce many of these signals in a short time: someone who has moved house, got a new phone and changed their number, forgotten their password and failed several logins, and is travelling. Each is innocent, and together they look like a takeover.
A house move, in particular, changes the device, the location and the contact details nearly at once. Reducing this means recognising customer-initiated changes confirmed through a trusted channel, and weighting signals that the customer has verified as lower risk than ones that have not.
Investigating an alert from this rule
- Build a timeline of the account from the earliest signal, in order: logins, failures, changes, enrolments, payee additions, payments.
- Establish whether the profile changes were verified through a channel that was already trusted before the change.
- Compare each session's device, location and network with the account's history.
- Look for money movement: what has already left, where it went, and whether anything can still be recalled.
- Contact the customer using a detail that predates the changes, not one introduced during the suspicious period.
- If takeover is confirmed, record what the attacker changed so it can be reversed, and tag the account so later activity is judged accordingly.
Related rules
Where this rule fits
The use case this rule supports is covered in account takeover, and the fraud rules engine page describes how conditions like these are expressed and versioned.
Common questions
What are account takeover indicators in a fraud rule?
Account takeover indicators are the signals that together suggest someone other than the owner controls an account: a new device, a change of credentials or contact details, an unusual location, repeated failed logins, and a first payment to a new payee. No one signal is conclusive, so a rule combines several and weighs them.
The value is in the combination and the order. A contact detail change followed shortly by a payment to a new payee is far more telling than either alone, so rules usually weight the signals and require a minimum number.
Genuine customers also produce several of these signals at once, most obviously when they move house, so confirmed customer-initiated changes are weighted lower than unconfirmed ones.
Discuss your transaction monitoring requirements
Tell us which transaction flows you need to monitor, which systems Akoben would integrate with, and how your fraud team works today.
Request a demo