Java/Kotlin lightweight implementation of RFC-6238 and RFC-4226 to generate and validate time-based one-time passwords (TOTP).
- Get an invite to the Slack channel by filling in this form.
Check the latest package at Maven central on: https://mvnrepository.com/artifact/com.atlassian/onetime.
<dependency>
<groupId>com.atlassian</groupId>
<artifactId>onetime</artifactId>
<version>2.1.1</version>
</dependency>
implementation("com.atlassian:onetime:2.1.1")
libraryDependencies += "com.atlassian" % "onetime" % "2.1.1"
This library is suitable for both prover and verifier.
As described in RFC-6238, the prover is the user trying to authenticate and submitting the generated TOTP to the verifier.
An example is: The user is trying to access an online service and is required to input the TOTP after authenticating with username and password. The user will need an application to generate such TOTP.
To generate the current TOTP for a given user's secret, do:
Assuming the user's enrollment has a base 32 encoded secret key: ZIQL3WHUAGCS5FQQDKP74HZCFT56TJHR
We can then generate the TOTP for the current time step:
val secret = TOTPSecret.fromBase32EncodedString("ZIQL3WHUAGCS5FQQDKP74HZCFT56TJHR")
val totpGenerator: TOTPGenerator = TOTPGenerator()
val totp = totpGenerator.generateCurrent(secret) //TOTP(value=123456)
You can now present the value of totp
to the user.
As described in RFC-6238, the verifier is the system validating the user's credentials, including multi-factor authentication. After the system validates the username/password, it then verifies the user's TOTP input.
But before being able to verify a user's TOTP, we need to enrol the user.
Upon an enrolment request, the system will
- Generate a secret and store it for the user and other information the system deems necessary.
- Generate a TOTP URI for further QR code generation that will allow the user to enrol with mobile apps such as Google Authenticator
This library provides two out of the box secret generators:
- ASCII range secret generator
- Random secret generator
But it is not limited to that. If you want to use a more complex secret generator using, for example GCP or AWS, you'll benefit from implementing a non-blocking secret generator.
1time provides 3 examples for this sake:
- Blocking generator
- Generator that returns
CompletableFuture
- Generator using Kotlin coroutines and thus relying on continuations (CPS)
fun interface SecretProvider {
fun generateSecret(): TOTPSecret
}
This is the simplest generator that doesn't rely on any complex generator. We provide two examples ready to use, ASCII and random:
class AsciiRangeSecretProvider : SecretProvider {
companion object {
private val ASCII_RANGE: CharRange = (' '..'z')
fun generateSecret() = TOTPSecret(
(1..20).map { ASCII_RANGE.random() }.joinToString("").toByteArray()
)
}
override fun generateSecret() = AsciiRangeSecretProvider.generateSecret()
}
class RandomSecretProvider : SecretProvider {
companion object {
fun generateSecret() =
SecureRandom().let {
val byteArray = ByteArray(20)
it.nextBytes(byteArray)
TOTPSecret(byteArray)
}
}
override fun generateSecret() = RandomSecretProvider.generateSecret()
}
To use any of these, just do:
val secret = RandomSecretProvider.generateSecret()
This is particularly useful when dealing with a complex, external resource for better random generation and when using Java.
fun interface AsyncSecretProvider {
fun generateSecret(): CompletableFuture<TOTPSecret>
}
We provide examples of these too, e.g.:
class AsyncAsciiRangeSecretProvider : AsyncSecretProvider {
override fun generateSecret(): CompletableFuture<TOTPSecret> =
CompletableFuture.supplyAsync {
AsciiRangeSecretProvider.generateSecret()
}
}
To use any of these, just do:
val secret: CompletableFuture<TOTPSecret> = AsyncAsciiRangeSecretProvider().generateSecret()
//handle completable future
If you are using Kotlin, this should be the default approach. Even if dealing with external random generator, it will be non-blocking by default:
fun interface CPSSecretProvider {
suspend fun generateSecret(): TOTPSecret
}
We provide examples of these too:
class CPSAsciiRangeSecretProvider : CPSSecretProvider {
override suspend fun generateSecret(): TOTPSecret =
AsciiRangeSecretProvider.generateSecret()
}
To use any of these, just do:
runBlocking {
val secret: TOTPSecret = CPSAsciiRangeSecretProvider().generateSecret()
// use secret
}
Or in your suspend
functions without blocking.
val totpUri = service.generateTOTPUrl(
secret, ////NIQXUILREVGHIUKNORKHSJDHKMWS6UTY
EmailAddress("[email protected]"),
Issuer("Acme Co")
)
totpUri // otpauth://totp/Acme+Co:jsmith%40acme.com?secret=NIQXUILREVGHIUKNORKHSJDHKMWS6UTY&issuer=Acme+Co&algorithm=SHA1&digits=6&period=30
Once generated the totpUrl, the system can use any library of choice to generate the TOTP QR code. For example:
After the user has successfully enrolled, both the user and the system will share the secret.
When the user (prover) is trying to authenticate, the system (verifier) will validate the user's TOTP input.
In order to do so, do:
val userInput: TOTP = TOTP("123456") //TOTP from user input
val result = service.verify(
userInput,
secret //NIQXUILREVGHIUKNORKHSJDHKMWS6UTY
)
The obtained result
can be either
TOTPVerificationResult.InvalidOtp
: The provided TOTP is invalid for the secret and default configurationTOTPVerificationResult.Success
: The provided TOTP is valid for the secret and default configuration. In addition, the index of the matching TOTP is provided in theSuccess
data class.
When the verification is successful, the type of result
will be TOTPVerificationResult.Success
. This data class contains an index value:
data class Success(val index: Int) : TOTPVerificationResult()
The index value indicates the window on which the TOTP was found. More on windows here.
When the verifier allows, for instance, 1 past window, one present window (mandatory) and one future window, upon successful verification, the index value will be:
- -1 when successful on past window or T - 1
- 0 when successful on current window or T
- 1 when successful on future window or T + 1
Where T is the time step. This is calculated by current epoch time in seconds divided by (floor) time step size (30 seconds by default).
This index is useful if your solution wants to implement resynchronization.
From RFC-6238:
Upon successful validation, the validation server can record the detected clock drift for the token in terms of the number of time steps. When a new OTP is received after this step, the validator can validate the OTP with the current timestamp adjusted with the recorded number of time-step clock drifts for the token.
By keeping track of the successful indexes per user device, the verifier could:
- Adjust the allowed windows to cater for prover's clock drifts.
- Notify the user that the device's clock is drifting and could end up in blocking access to the account, asking the user to re-enrol.
The previous instructions rely on a set of defaults.
TOTP Generator constructor takes the following parameters and corresponding default values:
val totpGenerator: TOTPGenerator = TOTPGenerator(
clock = Clock.systemUTC(),
startTime = 0,
timeStepSeconds = 30,
otpLength = OTPLength.SIX,
digest = HMACDigest.SHA1
)
startTime
: Epoch second from which the counter starts. By default is 0, which points to Thursday, January 1, 1970 0:00:00 GMT.timeStepSeconds
: The duration of a TOTP. By default, is 30 seconds.otpLength
: TOTP digit count. By default, is 6 digits.digest
: HMAC algorithm. By default, is SHA1.
See Google Authenticator handling of these parameters: Key-Uri-Format
These defaults apply for the verifier.
TOTP Service constructor takes the following parameters and corresponding default values:
val service = DefaultTOTPService(
totpGenerator = TOTPGenerator(),
totpConfiguration = TOTPConfiguration()
)
totpGenerator
: The default generator as described before. You could create different instances ofDefaultTOTPService
with different configurations oftotpGenerator
. This may be useful if your solution stores detailed configuration such as TOTP length, time step size and hashing algorithm per enrolment instead of a global configuration. If enrolment will allow the user to choose these parameters, then make sure the prover app supports such configuration.totpConfiguration
: This configuration allows to tune how the verification behaves in terms of accepted windows and how to generate secret keys.secretProvider
: By default we provideAsciiRangeSecretProvider
. This is a random key generator that generates 20 bytes in the ASCII range. This is useful for when you need the raw secret value to be composed of printable characters. You can provide your own Secret provider by extendingSecretProvider
functional interface.allowedPastSteps
: By default, this is 0. As specified by RFC-6238, the verifier should accept the TOTP from the previous time step to cater for possible network delays. If you want to accept past windows as valid, increase this number.allowedFutureSteps
: By default, this is 0. Same asallowedPastSteps
but to cater for possible clock drifts in which the prover's clock is slightly in the future relative to the verifier's. If you want to accept future time windows as valid, increase this number. More on time windows on the next section.
When both the prover's and verifier's clocks are in sync, this is what it would look like when the verifier allows one past window and 1 future window:
The current TOTP for the prover is what the server sees in the middle of the window, since it now accepts TOTPs from 3 time windows; current window, the previous and the next.
But when the prover's clock is delayed relative to the verifiers, this is what it would look like:
When the verifier generates the list of 3 TOTPs for the user, the one the prover provided is already in the past.
This can also happen due to network delay. Say the prover submitted the TOTP close to the end of the time window of 30 seconds and given network delays, upon calculation on the verifier, that TOTP value is no longer the current, but the previous. In these cases it makes sense to deem that TOTP as valid to prevent the prover from having to resubmit the TOTP.
Similarly to the prover's clock being delayed, it can also happen that it is slightly in the future relative to verifier's clock. The opposite to the diagram above would happen and upon calculation on verifier's side, the TOTP provided will be the next to current. It also makes sense to deem that attempt as valid.
Contributions to 1time are welcome! Please see CONTRIBUTING.md for details.
Copyright (c) 2022 Atlassian and others. Apache 2.0 licensed, see LICENSE file.