Detecting Account Takeover

Account takeover is the hijacking of a genuine customer's account by someone who then operates it as the customer. It is among the most costly fraud patterns because the account has a real history, real balance and real trust, and all three are available to the attacker.

What this fraud pattern looks like

A typical takeover has a recognisable shape. The attacker obtains credentials, through phishing, a breach of another service, malware on the customer's device, or by taking over the customer's phone number in a SIM swap. They sometimes try the credentials repeatedly first, since lists of stolen passwords are tested against many accounts. Once in, they usually act quickly: they add a new beneficiary, sometimes change the contact details or register a new device, and move money out, often in a few payments close together.

The account then looks, in sequence, like this: a failed run of attempts, a successful login from an unfamiliar device, a new payee, and a large transfer or several smaller ones. In a wallet, the equivalent is a transfer followed by a cash-out.

Why it is hard to detect

The central difficulty is that authentication succeeded. The credentials were correct, so every control that asks whether this is the account holder has already answered yes. A password check cannot object to a correct password. A one-time code sent to a number the attacker controls, after a SIM swap, is also correct. The controls are working as designed; they have simply been handed the right answers.

That moves the whole detection problem downstream, to behaviour after login, where there is much less certainty. A new device is not proof: people buy phones. A new beneficiary is not proof: people pay new people. Each signal has a large innocent population, so a rule on any one of them either misses the attack or floods analysts with alerts about genuine customers.

Speed makes it worse. The window between a takeover beginning and the money leaving can be short, and the attacker has often studied the account enough to keep amounts plausible. And because the account's history is genuine, a baseline built from it describes the holder, which only helps if the attacker behaves differently from the holder, as they usually do, in device, pace and the way they navigate.

The signals that expose it

Detection reads a set of weak signals as a sequence:

Order matters as well as presence. A new device followed by a new payee followed by a large transfer inside a short window is the signature, and the elapsed time between the steps is part of it.

The rules that catch it

Because no single signal is conclusive, the rules for takeover are narrow checks that are most useful in combination:

A composite works because it demands that independent weak signals coincide. The probability that a genuine customer produces a new device, a new payee and a near-total transfer within minutes is far lower than for any one of those events, and the rule can be tuned on the window and the combination rather than on a single threshold.

Investigating it

Where the account has been confirmed as compromised, the response is usually a combination of reversing what can be reversed, locking the credentials, and re-establishing control of the account through a channel the attacker does not hold. Many institutions also raise the authentication requirement on the account for a period afterwards — a step-up authentication challenge on sensitive actions — on the reasoning that an account which has been taken over once is a more attractive target than one which has not.

An analyst assembling a takeover case reconstructs the sequence: when the failures began, when the new device first appeared, when the beneficiary was added, and how long the gap was between each step and the payment. Elapsed time is the most informative fact on the page, because genuine customers rarely move that quickly through a new device and a new payee.

Next they establish what the holder knows. Contacting the customer through a channel the attacker does not control, not one that may have been changed, is the surest test. They also check for changes to contact details, since an attacker who has changed them may be intercepting messages. Finally they trace where the funds went: a beneficiary receiving money from several hijacked accounts is probably a mule account, and that finding is worth more than the single case.

How Akoben handles it

Akoben receives account and transaction events asynchronously, over REST or as batch files, and enriches each with the account's recent history. It compares activity with a behavioural baseline built from that account's own history and evaluates it against the active ruleset, so the combinations described above can be expressed as rules. A match raises an alert, which Akoben routes into a case queue for assignment, escalation and SLA tracking.

Rules can be changed by an analyst without a release. Each change creates a new ruleset, applied as a whole and reverted to the previous version if needed. A ruleset can be checked against historical transactions in batch, and run in alerting-only mode while a team judges the alert volume. A tag applied to a confirmed takeover carries forward, so later transactions on that account are scored differently. Rule-effectiveness and false-positive-rate reports show whether each takeover rule is finding fraud or only generating work.

See the fraud rules engine for how rules are written and versioned, and case management for the investigation workflow.

Common questions

What is account takeover fraud?

Account takeover fraud is when a criminal gains control of a genuine customer's account and uses it to move money or change its details. The attacker typically obtains the credentials through phishing, data breaches, malware or a SIM swap, then logs in as the customer. Because the login is real, the fraud shows up in what happens after authentication: new devices, new beneficiaries and unfamiliar transaction patterns.

Account takeover is a way of committing payment fraud rather than a separate kind of loss. The attacker's goal is usually to drain the balance, often by adding a beneficiary they control and sending funds to it.

Read the glossary definition of account takeover

How is account takeover detected?

Account takeover is detected by watching what an account does after login, because the login itself looks genuine. Monitoring compares activity with the account's history and looks for a new device, a beneficiary added shortly before a payment, a run of failed attempts, a change of contact details, and transfers that drain the balance. Several of these together, in a short window, are far stronger evidence than any one alone.

Detection therefore rests on sequence and combination. A new device is common and a new beneficiary is common; a new device, then a new beneficiary, then a near-total transfer within minutes is a pattern few genuine customers produce.

See the account takeover indicator rule

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