Home

Awesome

jsonwebtoken

API documentation on docs.rs

See JSON Web Tokens for more information on what JSON Web Tokens are.

Installation

Add the following to Cargo.toml:

jsonwebtoken = "9"
# If you do not need pem decoding, you can disable the default feature `use_pem` that way:
# jsonwebtoken = {version = "9", default-features = false }
serde = {version = "1.0", features = ["derive"] }

The minimum required Rust version (MSRV) is specified in the rust-version field in this project's Cargo.toml.

Algorithms

This library currently supports the following:

How to use

Complete examples are available in the examples directory: a basic one and one with a custom header.

In terms of imports and structs:

use serde::{Serialize, Deserialize};
use jsonwebtoken::{encode, decode, Header, Algorithm, Validation, EncodingKey, DecodingKey};

/// Our claims struct, it needs to derive `Serialize` and/or `Deserialize`
#[derive(Debug, Serialize, Deserialize)]
struct Claims {
    sub: String,
    company: String,
    exp: usize,
}

Claims

The claims fields which can be validated. (see validation)

#[derive(Debug, Serialize, Deserialize)]
struct Claims {
    aud: String,         // Optional. Audience
    exp: usize,          // Required (validate_exp defaults to true in validation). Expiration time (as UTC timestamp)
    iat: usize,          // Optional. Issued at (as UTC timestamp)
    iss: String,         // Optional. Issuer
    nbf: usize,          // Optional. Not Before (as UTC timestamp)
    sub: String,         // Optional. Subject (whom token refers to)
}

Header

The default algorithm is HS256, which uses a shared secret.

let token = encode(&Header::default(), &my_claims, &EncodingKey::from_secret("secret".as_ref()))?;

Custom headers & changing algorithm

All the parameters from the RFC are supported but the default header only has typ and alg set. If you want to set the kid parameter or change the algorithm for example:

let mut header = Header::new(Algorithm::HS512);
header.kid = Some("blabla".to_owned());
let token = encode(&header, &my_claims, &EncodingKey::from_secret("secret".as_ref()))?;

Look at examples/custom_header.rs for a full working example.

Encoding

// HS256
let token = encode(&Header::default(), &my_claims, &EncodingKey::from_secret("secret".as_ref()))?;
// RSA
let token = encode(&Header::new(Algorithm::RS256), &my_claims, &EncodingKey::from_rsa_pem(include_bytes!("privkey.pem"))?)?;

Encoding a JWT takes 3 parameters:

When using HS256, HS384, or HS512, the key is always a shared secret like in the example above. When using RSA/EC, the key should always be the content of the private key in PEM or DER format.

If your key is in PEM format, it is better performance wise to generate the EncodingKey once in a lazy_static or something similar and reuse it.

Decoding

// `token` is a struct with 2 fields: `header` and `claims` where `claims` is your own struct.
let token = decode::<Claims>(&token, &DecodingKey::from_secret("secret".as_ref()), &Validation::default())?;

decode can result in errors for a variety of reasons:

As with encoding, when using HS256, HS384, or HS512, the key is always a shared secret like in the example above. When using RSA/EC, the key should always be the content of the public key in PEM (or certificate in this case) or DER format.

In some cases, for example if you don't know the algorithm used or need to grab the kid, you can choose to decode only the header:

let header = decode_header(&token)?;

This does not perform any signature verification or validate the token claims.

You can also decode a token using the public key components of a RSA key in base64 format. The main use-case is for JWK where your public key is in a JSON format like so:

{
   "kty":"RSA",
   "e":"AQAB",
   "kid":"6a7a119f-0876-4f7e-8d0f-bf3ea1391dd8",
   "n":"yRE6rHuNR0QbHO3H3Kt2pOKGVhQqGZXInOduQNxXzuKlvQTLUTv4l4sggh5_CYYi_cvI-SXVT9kPWSKXxJXBXd_4LkvcPuUakBoAkfh-eiFVMh2VrUyWyj3MFl0HTVF9KwRXLAcwkREiS3npThHRyIxuy0ZMeZfxVL5arMhw1SRELB8HoGfG_AtH89BIE9jDBHZ9dLelK9a184zAf8LwoPLxvJb3Il5nncqPcSfKDDodMFBIMc4lQzDKL5gvmiXLXB1AGLm8KBjfE8s3L5xqi-yUod-j8MtvIj812dkS4QMiRVN_by2h3ZY8LYVGrqZXZTcgn2ujn8uKjXLZVD5TdQ"
}
// `token` is a struct with 2 fields: `header` and `claims` where `claims` is your own struct.
let token = decode::<Claims>(&token, &DecodingKey::from_rsa_components(jwk["n"], jwk["e"]), &Validation::new(Algorithm::RS256))?;

If your key is in PEM format, it is better performance wise to generate the DecodingKey once in a lazy_static or something similar and reuse it.

Convert SEC1 private key to PKCS8

jsonwebtoken currently only supports PKCS8 format for private EC keys. If your key has BEGIN EC PRIVATE KEY at the top, this is a SEC1 type and can be converted to PKCS8 like so:

openssl pkcs8 -topk8 -nocrypt -in sec1.pem -out pkcs8.pem

Validation

This library automatically validates the exp claim, and nbf is validated if present. You can also validate the sub, iss, and aud but those require setting the expected values in the Validation struct. In the case of aud, if there is a value set in the token but not in the Validation, the token will be rejected.

Validation is only made on present fields in the claims. It is possible to define the required claims, hence verifying that a JWT has a value for each of these claims before it is considered for validation. The required claims can be set in the Validation struct. The default option requires the exp claim to be present.

Since validating time fields is always a bit tricky due to clock skew, you can add some leeway to the iat, exp, and nbf validation by setting the leeway field.

Last but not least, you will need to set the algorithm(s) allowed for this token if you are not using HS256.

Look at examples/validation.rs for a full working example.