Smart Signals Overview
Fingerprint is not just an identity platform; it also helps fight fraud by providing actionable device intelligence called Smart Signals. Read on to get a basic understanding of individual Smart Signal components and their applications.
Setup and Usage
Typical integration into a project consists of those steps:
- Add the JavaScript agent on the frontend part of your web application. The smart signals are collected during the
load()
function. If you are an existing customer and already deployed Fingerprint, you can skip this step. - Get the
requestId
from the JavaScript agent response and pass it to your server. - Call GET /events endpoint from your server-side environment to retrieve the Smart Signals in the API response.
- Use the information about fraud in your business logic.
Availability
Because Smart Signals are focused on fraud detection, we only make them available through our Server API, Webhooks, and Sealed Client Results so we can ensure you receive those results unaltered.
Smart Signals are available in our Pro Plus and Enterprise plans. If you are an Enterprise plan customer interested in one or more of these signals, please contact our support team to enable them for you. Signing up for an Enterprise plan does not automatically enable these signals.
Suspect Score
Suspect Score combines all Smart Signals into a weighted integer value based on their global probability. You can read more about the weights and the mechanisms behind Suspect Score in a dedicated section.
"suspectScore": {
"data": {
"result": 0
}
}
Velocity Signals
Velocity Signals sum key data points for a specific visitorID
at three distinct intervals: 5 minutes, 1 hour, and 24 hours as follows:
- Number of unique IP addresses linked to the
visitorID
- Number of unique countries associated with a
visitorID
- Number of unique
linkedIDs
associated with thevisitorID
- Number of events attributed to the
visitorID
For high-activity visitors exceeding a count of 20,000 within a 24-hour window, we will omit the aggregation of 24-hour results for unique IP addresses, countries, and LinkedIDs.
{
"velocity": {
"data": {
"distinctIp": {
"intervals": {
"5m": 1,
"1h": 1,
"24h": 5
}
}
}
}
}
{
"velocity": {
"data": {
"distinctCountry": {
"intervals": {
"5m":1,
"1h": 1 // For visitors exceeding 20,000 in 24 hours, we omit 24-hour aggregations
}
}
}
}
}
{
"velocity": {
"data": {
"distinctLinkedId": {} // This will be an empty object in case there is no data to present
}
}
}
{
"velocity": {
"data": {
"events": {
"intervals": {
"5m": 3,
"1h": 5,
"24h": 23
}
}
}
}
}
Browser Smart Signals
Each individual Smart Signal contributes with a unique piece of information about the device. Most of our Smart Signals are independent from the source of the request (browser vs. native mobile application) but there are some signals that are not available if the data was collected through our official mobile SDKs.
See the tags under each title or our Smart Signals Cheat Sheet to determine whether you should expect to see the Smart Signal output in Server API and Webhooks based on its source.
Browser Bot Detection
Get information about good and bad bots, allowing you to block or filter them from your regular traffic and stop automated abuse attempts.
The bot.result
field contains one of notDetected
, bad
or good
values where
notDetected
means that we haven't detected a bot and the payload should belong to a regular visitorgood
indicates that the bot is a well-known web crawler or other search engine botbad
is an automated tool that doesn't have legitimate uses and assumes fraudulent activity
"botd": {
"data": {
"bot": {
"result": "bad",
"type": "headless"
},
"url": "https://example.com/",
"ip": "193.165.141.254",
"time": "2024-04-12T09:29:59.588Z",
"userAgent": "Headless",
"requestId": "1712914199539.K1EXmu"
}
}
"botd": {
"data": {
"bot": {
"result": "notDetected"
},
"url": "https://example.com/",
"ip": "94.142.239.124",
"time": "2023-08-29T21:48:17.351Z",
"userAgent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/116.0.0.0 Safari/537.36",
"requestId": "1693345697334.YKOdT5"
}
}
Browser Incognito Detection
Incognito flag detects whether incognito or private modes are being used by the visitor to access the customer's website.
Most browsers have an option called Incognito mode (Chrome) or Private mode (Safari, Firefox) that disables some functionality and makes the general browser storage more ephemeral. Normal browser tabs save additional information like cookies and local storage content in a way that ensures that the information is available to website even after the browser window is closed and re-opened. Incognito mode makes sure that there are no such traces once a window gets closed.
Since the most basic fingerprinting technique could be storing a cookie in the visitor's browser, the incognito mode can be used to circumvent this simple mechanism, making sure that the returning visitor is not correctly re-identified through cookies and other information stored in the browser. This might be interpreted in an attempt to pose as a completely new identity. It might also be considered to be the most basic mechanism of browser tampering.
The flag is an interesting data point in fraud scenarios, especially in use cases where applying promotional coupons, redeeming bonuses or voting should be limited to once per user.
"incognito": {
"data": {
"result": true
}
}
"incognito": {
"data": {
"result": false
}
}
These browsers are supported:
- Safari (macOS / iOS)
- Chrome (Windows / macOS / iOS / Android / Linux)
- Samsung Browser (Android)
- Firefox (Windows / macOS / iOS / Android / Linux)
- Edge (Windows / macOS / iOS / Android)
- Brave (Windows / macOS / iOS / Android / Linux)
HTML iframes limitation
Incognito detection can give a false result when JS agent runs in an iframe. Make sure JS agent runs outside iframes to get the accurate result.
IP Geolocation
Provides information about the physical location of the originating IP address.
We deploy several mechanisms to detect the IP address of the original client which allows us to correctly determine visitor location, even when using anonymizing tools. The geolocation
field shows the estimated physical location of the client.
The asn
and datacenter
field can be used to group ranges of IP addresses that belong to the same owner to apply protection rules against a whole block of IP addresses. They could also indicate that the visitor connects through an intermediate server instead of connecting through an ISP directly.
"ipInfo": {
"data": {
"v4": {
"address": "94.142.239.124",
"geolocation": {
"accuracyRadius": 20,
"latitude": 50.05,
"longitude": 14.4,
"postalCode": "150 00",
"timezone": "Europe/Prague",
"city": {
"name": "Prague"
},
"country": {
"code": "CZ",
"name": "Czechia"
},
"continent": {
"code": "EU",
"name": "Europe"
},
"subdivisions": [
{
"isoCode": "10",
"name": "Prague"
}
]
},
"asn": {
"asn": "7922",
"name": "COMCAST-7922",
"network": "73.136.0.0/13"
},
"datacenter": {
"result": true,
"name": "DediPath"
}
},
"v6": {
"address": "::ffff:5e8e:ef7c",
"geolocation": {
"accuracyRadius": 20,
"latitude": 50.05,
"longitude": 14.4,
"postalCode": "150 00",
"timezone": "Europe/Prague",
"city": {
"name": "Prague"
},
"country": {
"code": "CZ",
"name": "Czechia"
},
"continent": {
"code": "EU",
"name": "Europe"
},
"subdivisions": [
{
"isoCode": "10",
"name": "Prague"
}
]
},
"asn": {
"asn": "7922",
"name": "COMCAST-7922",
"network": "::ffff:5e8e:ef7c/110"
},
"datacenter": {
"result": true,
"name": "DediPath"
}
}
}
}
VPN Detection for browsers
Fingerprint's VPN detection is capable of detecting whether the user is in a different timezone compared to their originating IP address (timezoneMismatch
field). It can also provide information about whether an IP address has been known to belong to one of the public VPN providers through the publicVPN
field. On top of that, we detect whether the OS of the client matches the OS of the network traffic we see, represented in the osMismatch
flag.
If either of those detection methods
return positive results, we return the final result
as true
.
"vpn": {
"data": {
"result": true,
"originTimezone": "Europe/Prague",
"originCountry": "unknown" // Not yet supported for browsers
"methods": {
"timezoneMismatch": true,
"publicVPN": true,
"auxiliaryMobile": false, // Irrelevant for browsers, mobile SDKs only
"osMismatch": true
}
}
}
"vpn": {
"data": {
"result": false,
"originTimezone": "Europe/Prague",
"originCountry": "unknown" // Not yet supported for browsers
"methods": {
"timezoneMismatch": false,
"publicVPN": false,
"auxiliaryMobile": false, // Irrelevant for browsers, mobile SDKs only
"osMismatch": false
}
}
}
result
result
true
if we detect a VPN throughtimezoneMismatch
,publicVPN
, orosMismatch
methods. The field can be used as a standalone flag for VPN detection. Alternatively, fields in themethods
section (documented below) can be used in specific use-cases to improve the granularity of the detectionfalse
if none of themethods
below returned a positive result
originTimezone
originTimezone
This field exposes the IANA timezone that is available through the system settings. Since the presence of VPN does not affect the value of this field, it can be used as a true timezone of origin (physical location) of the user.
- Example:
originTimezone: "Europe/Prague"
originCountry
originCountry
This field is populated only when the request came from our iOS and Android native SDKs and is always set to "unknown"
for browsers. See the detailed description in VPN Detection for mobile devices .
methods
methods
More granular VPN detection methods that contribute to the final VPN detection result
.
timezoneMismatch
timezoneMismatch
true
when we detect that the geolocated IP address of the device does not matchoriginTimezone
read from the system settingsfalse
whenoriginTimezone
matches the location of the IP address from which we receive the traffic
The originTimezone
includes a country-specific reference like "Europe/Prague"
. The timezoneMismatch
will be true
even if a visitor in Prague uses a Berlin VPN server, despite both countries being in the same GMT+2 UTC offset. For most use cases, timezoneMismatch
functions well as an indicator of country mismatch between the browser and the IP address.
publicVPN
publicVPN
Detects whether the IP address of the event matches any publicly known VPN providers like NordVPN, Surfshark , ExpressVPN, ProtonVPN and others.
true
when our database of public VPN provider IP addresses contains the IP address of the eventfalse
if we did not find a match in our public VPN provider database
osMismatch
osMismatch
Compares the network signature of the request we have received through our proprietary algorithm and searches our internal database to determine whether the signature matches the OS (operating system) of the client.
true
if the network signature of the request we receive does not match the signature of the OS of the clientfalse
- if the network signature of the request we receive matches the signature of the OS of the client
- if we fail to capture the OS information through our detection algorithm (happens in 10-15% cases)
The method is currently very accurate for Windows devices but might be less accurate for other OS combinations. We recommend using this flag to lower the false positive rate of the timezoneMismatch
method.
auxiliaryMobile
auxiliaryMobile
This field is populated only when the request came from our iOS and Android native SDKs and is always set to false
for browsers. See the detailed description in VPN Detection for mobile devices .
Browser Tamper Detection
There are very simple techniques to confuse less sophisticated fingerprinting algorithms through User Agent spoofing or trying to change the output of selected signals that are collected from the browser.
While our visitorID
remains stable in those cases, tampering detection helps detect this behavior and flag visitors that have tried applying such techniques by comparing their browser signature to our statistical model.
The output of this model is captured as anomalyScore
, a number indicating how improbable is the signature of the visitor's browser. Values close to 1
signify highly anomalous browsers and we consider anything above the threshold of 0.5
to be actionable (the result
field conveniently captures that fact).
"tampering": {
"data": {
"result": true,
"anomalyScore": 0.97 // close to 1 = high level of confidence
}
}
"tampering": {
"data": {
"result": false,
"anomalyScore": 0
}
}
Virtual Machine Detection
Detects if the browser is running inside a virtualization software like VirtualBox by examining the browser configuration.
"virtualMachine": {
"data": {
"result": true
}
}
"virtualMachine": {
"data": {
"result": false
}
}
Privacy-Focused Settings
Firefox, Brave, and other privacy focused browsers actively fight fingerprinting and expose user settings that have the ability to randomize and obfuscate signal output. The Privacy-Focused Settings Smart Signal detects if those settings are enabled and reports it.
"privacySettings": {
"data": {
"result": true
}
}
"privacySettings": {
"data": {
"result": false
}
}
Developer Tools Detection
Detect if developer tools are manually opened in Chrome or Firefox browsers.
"developerTools": {
"data": {
"result": true
}
}
"developerTools": {
"data": {
"result": false
}
}
Remote Control Tools Detection
To detect the usage of remote control tools by checking if the request originated from a device being remotely controlled or remotely controlling another device, you must first enable this feature. Data will not be collected unless the remoteControlDetection
flag is set to true
. Once enabled, please contact our support team to activate the result exposure.
The currently supported remote tools are:
- AnyDesk: Supported on Windows, macOS, and Android (Webview)
- TeamViewer and RDP: Supported on Windows
"remoteControl": {
"data": {
"result": true
}
}
"remoteControl": {
"data": {
"result": false
}
}
IP Blocklist Matching
Block IP addresses based on their presence in different public and proprietary blocklists.
The feature uses a combination of our and vendor IP databases to determine whether
- The IP address is a known tor exit node (
tor
payload) - The IP address has been part of a network attack or email spam attack (
ipBlocklist.details
payload) - The IP address belongs to a public proxy provider (
proxy
payload)
tor
andproxy
payloads do not utilize active Tor or proxy detection mechanisms.
"ipBlocklist": {
"data": {
"result": false,
"details": {
"emailSpam": false,
"attackSource": false
}
}
}
"tor": {
"data": {
"result": false
}
}
"proxy": {
"data": {
"result": false
}
}
High-Activity Device
High-Activity Device searches for spikes in traffic connected to a single visitorID
. It computes the amount of visits in the past 24 hours for each visitorID
every five minutes and reports data.result
as true
if the number of visits for the past 24 hours surpasses a threshold.
High-Activity Device Threshold
The threshold is calculated for each customer separately and is determined through a normal traffic distribution. If the returned visitorID
does more requests than 98% of other daily visitors, the High-Activity Smart Signal returns a positive result. If the traffic doesn't pass the threshold, data.result
returns false
and dailyRequests
field isn't provided.
"highActivity": {
"data": {
"result": true,
"dailyRequests": 42
}
}
"highActivity": {
"data": {
"result": false
}
}
Raw Device Attributes
Raw Device Attributes expose additional data points we normally collect through our JavaScript agent. This Smart signal is only available for web browsers.
Format Stability
Raw Device Attributes contents can change at any time in the future (includes complete scoped field removal) and we advise against setting dependencies on concrete field names inside of the
data
field.
"rawDeviceAttributes": {
"data": {
"<field_name>": {
"value": 127
}
}
Smart Signals for mobile devices
The smart signals listed below are applicable only to mobile devices. These signals provide valuable intelligence about the integrity of the device as well as the app. These smart signals are available as part of our Pro Plus and Enterprise plans. If you are interested in one or more of these signals, then please contact our support team to enable them for you.
When your app is integrated with Fingerprint (using either our native mobile SDKs or our multi-platform SDKs), you will start receiving these signals as part of our Server API and/or Webhooks responses.
Android Emulator Detection
Prevent spam and protect against nefarious Android emulator farms by ensuring the request is coming from a physical device. To detect emulators, several additional device attributes have to be collected and this increases the time to get a visitorId
. The latency can be anywhere between 300 and 1500ms.
"emulator": {
"data": {
"result": false
}
}
Android Tamper Detection
Ensure a safe Android mobile application environment by detecting rooted devices.
"rootApps": {
"data": {
"result": false
}
}
Cloned App Detection
Identify if a request is coming from a cloned application. Making multiple clones of an app on the same device can be used to abuse promotions and get around app restrictions. (Android only)
"clonedApp": {
"data": {
"result": false
}
}
Factory Reset Detection
Resetting a mobile device to default factory settings has been a common technique that fraudsters use to evade fraud detection mechanisms. When a factory reset is performed, it can
- Disguise the identity of the mobile device
- Be used to abuse promotional offers
- Prevent the mobile device from being blocked for fraudulent activity
The Factory Reset Smart Signal will indicate when the mobile device was most recently factory reset. You can use this information in your fraud detection mechanisms to flag a mobile device as suspicious.
As an example, for a mobile device that was factory reset on July 19, 2023, at 6 AM EST, the response will be:
"factoryReset": {
"data": {
// The time when the mobile device was factory reset in UTC
"time": "2023-07-19T011:00:00Z",
// The UTC 'time' represented as Unix epoch time
"timestamp": 1689445704
}
}
For mobile devices (and browsers) where a factory reset could not be detected, the time and timestamp fields in the response will correspond to the epoch as:
"factoryReset": {
"data": {
"time": "1970-01-01T00:00:00Z",
"timestamp": 0
}
}
Can a factory reset go undetected?
Yes. Here are a few possible scenarios in which a factory reset cannot be detected:
- The iOS device has genuinely not been factory reset since it was first activated.
- The iOS device that initiated the request is not a physical device, but an iOS simulator. (Conversely, a factory reset can be detected for Android emulators).
- Your iOS app does not use Fingerprint Identification SDK v2.3.0 or higher. The ability to detect factory reset was introduced only in v2.3.0.
- Your Android app uses an incompatible version of Fingerprint Identification SDK
- < v2.3.0: The ability to detect factory reset was introduced only in v2.3.0.
- v2.3.0 ≥ SDK < 2.3.4: Due to lazy signal collection, more than one request must be initiated from the same device for a factory reset to be detected. If you would like to use one of these versions in your app, you first need to contact our support team to enable factory reset detection for your account.
- ≥ v2.3.4: Since the first request from a fraudulent device is the most critical request to detect fraud, we made improvements that allowed us to detect a factory reset during the first request itself.
Frida Detection
In simple terms, app hooking is a process where the behavior of a mobile app is changed by intercepting the app at runtime. Frida is a dynamic instrumentation toolkit that tech-savvy fraudsters can use to spoof an app at runtime and change its behavior. With this smart signal, you will know if Frida is being used to dynamically instrument the app.
"frida": {
"data": {
"result": false
}
}
Geolocation Spoofing Detection
Fraudsters spoof the location of their mobile devices to fool fraud detection systems. With this smart signal, you will know if the location of the mobile device has been spoofed or not. For iOS only, this signal requires the host app to be granted Location Permission.
"locationSpoofing": {
"data": {
"result": false
}
}
Jailbroken Device Detection
An iPhone or an iPad is labeled jailbroken when unauthorized modifications are made to the underlying operating system (iOS). Jailbreaking weakens a device’s built-in security features and makes it easy to steal sensitive information stored on the device. See Unauthorized modification of iOS for more information.
Fraudsters employ phishing mechanisms to trick users into installing jailbreak software on their devices and then take control of their devices. Your app (e.g. banking, e-commerce, etc) will not be able to distinguish the transactions made using jailbroken devices from those made using genuine devices. Honoring these fraudulent transactions will lead to financial losses (caused by chargeback expenses, account takeover, etc) and loss of trust among your customers.
The Jailbroken Smart Signal will help you detect if your app is being used from an iPhone/iPad that has been jailbroken and enable you to proactively act upon it.
"jailbroken": {
"data": {
"result": false
}
}
VPN Detection for mobile devices
VPN apps for mobile devices are popular with both legit users and fraudsters. These VPN apps
- Protect their users’ privacy by encrypting all their traffic
- Allow their users to remain anonymous by giving them the ability to hide their real location
The VPN Detection Smart Signal for mobile devices will help you detect if your app is being used by a mobile device with an active VPN connection.
"vpn": {
"data": {
"result": false,
"originTimezone": "Europe/Berlin",
"originCountry": "DE" // country code string in the ISO 3166 format
"methods": {
"timezoneMismatch": false,
"publicVPN": false,
"auxiliaryMobile": false,
"osMismatch": false // always false, irrelevant for native SDKs
}
}
result
result
For mobile devices, result
is true when either of the four listed methods is true:
timezoneMismatch
-true
when the device is in a different timezone compared to the timezone of the originating IP address.publicVPN
-true
when the IP address matches one of the known IP addresses that belong to public VPN providers.auxiliaryMobile
-true
when other proprietary detection methods indicate the presence of a VPN.osMismatch
- alwaysfalse
when using mobile SDKs.
originTimezone
originTimezone
This field indicates the IANA timezone that is currently set for the mobile device. In situations where a VPN is enabled, this timezone can be different from the timezone associated with the IP address.
The originTimezone
usually includes a country-specific reference like "Europe/Prague"
. The timezoneMismatch
will be true
even if a mobile user in Prague uses a Berlin VPN server, despite both countries being in the same GMT+2 UTC offset. For most use cases, timezoneMismatch
functions well as an indicator of country mismatch between the mobile device and the IP address.
This field is supported for both Android and iOS devices.
originCountry
originCountry
This field indicates the country in which the mobile device is currently located. In situations where a VPN is enabled, this field can be very useful to know the real location of a device.
- For Android devices, this field is populated with the origin country only when your app is integrated with SDK v2.4.0 and higher. For lower versions of the SDK, the value is set to
unknown
. This field does not require Location Permissions to be enabled for your app. - For iOS devices,
originCountry
is not yet supported and the value is set tounknown
.
Updated about 11 hours ago