Geographic anomaly rule
A geographic anomaly rule detects activity from a location that is inconsistent with the account's established pattern, or with physical possibility. The rule compares each event's location with the account's history and flags a mismatch. Two common forms exist: activity from a place the account has never been associated with, and two events too far apart for one person to have travelled between in the elapsed time.
The fraud scenario this rule addresses
A card is cloned and used abroad while the owner is at home, or credentials are stolen and used from another country. In the impossible-travel form, a login or purchase in one city is followed minutes later by another in a different country, which a single person could not have done.
Location is also used for transactions where the point of sale is known, such as card payments and agent withdrawals, and a spread across distant places in a short time points to a compromised card.
Data the rule requires
- Location of each event, such as terminal location, IP-derived location or agent location
- Timestamp of each event
- The account's history of locations and home area
- Channel, since location accuracy differs between card terminals and IP addresses
- Known travel indicators where the customer has declared them
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 geographic_anomaly
FOR EACH event E on account A:
prev = most recent located event on A before E
distance = great_circle(E.location, prev.location)
elapsed = E.timestamp - prev.timestamp
implied_speed = distance / elapsed
impossible = implied_speed >= :max_plausible_speed
unfamiliar = E.location NOT IN usual_areas(A)
IF impossible OR (unfamiliar AND E.amount >= :min_amount):
RAISE alert(rule = "geographic_anomaly",
account = A, evidence = prev + E)
Choosing a threshold
The two forms have different parameters. For impossible travel the key one is the maximum plausible speed: set it close to commercial air travel and it catches only the clearest cases, set it lower and it alerts on a customer who has simply flown. For the unfamiliar-location form it is how an area is defined and how long a place stays familiar.
Location quality is a parameter in disguise. IP-derived location is coarse and often wrong, so a tight distance rule on it produces many false alerts, while terminal-based location is much more reliable. The same distance should not be treated the same way across channels.
Raising the amount floor on unfamiliar locations reduces noise but permits small test transactions abroad, which fraudsters use to confirm a card works before a larger one.
False positives to expect
Customers move: a holiday or business trip, a student abroad, a cross-border worker, someone using a VPN or a mobile network whose exit point is elsewhere, a corporate network that routes traffic through a distant data centre, and an online purchase from a merchant based in another country.
Travel cannot be predicted from transactions alone. Allowing customers to declare travel, treating a trip as a change of pattern once it has begun and been confirmed, and discounting IP-derived location where VPNs are common all help.
Investigating an alert from this rule
- Check the two events and the time between them, and confirm the distance and implied speed really are impossible.
- Establish how each location was derived and how reliable it is for that channel.
- Check whether a VPN, proxy or corporate network could explain the location.
- Compare the device and channel for each event: two different devices in two places points to two people.
- Look at whether other signs of travel exist, such as foreign currency purchases, a travel booking or a roaming indicator.
- Contact the customer through a known channel and ask where they are, then record the answer so a confirmed trip is not alerted on repeatedly.
Related rules
Where this rule fits
The use case this rule supports is covered in suspicious transactions, and the fraud rules engine page describes how conditions like these are expressed and versioned.
Common questions
What is a geographic anomaly rule?
A geographic anomaly rule is a fraud rule that raises an alert when activity comes from a location inconsistent with the account's pattern or with physical possibility. It has two common forms: activity from a place the account has never been associated with, and impossible travel, where two events are too far apart for one person to have covered in the time between them.
The reliability of the location data decides how the rule should be used. Terminal locations are precise, whereas IP-derived locations are coarse and are distorted by VPNs and corporate networks, so the same distance does not mean the same thing on every channel.
Holidays, business trips and cross-border workers produce the pattern legitimately, which is why customer-declared travel and confirmed trips are used to avoid repeat alerts.
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