How to Implement OAuth 2.0 PKCE Flow

Problem

Public clients (SPAs, mobile apps) cannot securely store a client secret, making them vulnerable to authorization code interception attacks. You need to implement PKCE (RFC 7636) to bind the authorization code to a code verifier known only to the client.

Prerequisites

• Understanding of OAuth 2.0 Authorization Code flow
• Familiarity with public vs. confidential clients
• Knowledge of SHA-256 hashing and Base64URL encoding

Steps

1. 1. Generate the code verifier and code challenge

   Generate a cryptographically random code verifier: a high-entropy string of 43-128 characters from the unreserved character set [A-Z][a-z][0-9]-._~. Then derive the code challenge: S256 method computes BASE64URL-ENCODE(SHA256(ASCII(code_verifier))). The code challenge is sent in the authorization request; the code verifier is kept locally and never sent until the token exchange. Use the random generator for the verifier and SHA-256 + Base64URL for the challenge.

   Input

   random 32 bytes -> base64url

   Output

   verifier: dBjftJeZ4CuK... | challenge: E9Melhoa2l2cYi8Yk... (S256)

2. 2. Send the authorization request with code challenge

   Redirect the user to the authorization endpoint with parameters: response_type=code, client_id, redirect_uri, scope, state (CSRF protection), code_challenge, code_challenge_method=S256. The authorization server stores the code challenge associated with the issued authorization code. The state parameter must be a random value that the client validates on redirect to prevent CSRF.

   Input

   GET /authorize?response_type=code&client_id=app1&code_challenge=E9Mel...&code_challenge_method=S256&state=xyz

   Output

   redirect: https://app.com/cb?code=SplxlOBeZQQYbYS6WxSbIA&state=xyz

3. 3. Exchange the code with the code verifier

   On receiving the authorization code, send a token request to the token endpoint with: grant_type=authorization_code, client_id, code, redirect_uri, code_verifier. The authorization server recomputes BASE64URL(SHA256(code_verifier)) and compares it to the stored code challenge. If they match, the code was not intercepted — issue tokens. If they don't match, reject the request. Use the OAuth token parser to inspect the returned tokens.

   Input

   POST /token { grant_type, code, code_verifier: dBjftJeZ4CuK... }

   Output

   { access_token, refresh_token, token_type: Bearer, expires_in }

4. 4. Validate tokens and secure the state parameter

   Validate the state parameter matches the one sent in step 2 to prevent CSRF. Parse the access token (if JWT) to check expiration, issuer, and audience. Store tokens securely: access token in memory, refresh token in secure storage (httpOnly cookie for web, secure enclave for mobile). Never store tokens in localStorage for SPAs. Use the JWT parser to verify token claims before using the token.

   Input

   access_token: eyJhbGciOi...

   Output

   valid: true | exp: 1234567890 | scope: read write

Expected result

A secure OAuth 2.0 PKCE flow where public clients authenticate without a client secret, protected against authorization code interception and CSRF attacks.

Common errors

Using plain code_challenge_method instead of S256

The 'plain' method sends the verifier directly as the challenge, providing no protection if intercepted. Always use S256 (SHA-256 based).

Code verifier too short or not cryptographically random

The verifier must be 43-128 characters of high entropy. Use a CSPRNG (crypto.getRandomValues), not Math.random(). 32 random bytes base64url-encoded gives 43 characters.

Omitting or not validating the state parameter

The state parameter prevents CSRF. Generate it randomly, include in the auth request, and validate it exactly matches on the redirect.

Troubleshooting

• If the authorization server rejects the token request with 'invalid_grant', verify the code_verifier matches the code_challenge used in the authorization request.
• For SPAs, consider using the Backend-For-Frontend (BFF) pattern instead of storing tokens in the browser, which is more secure than pure PKCE in the browser.
• If using refresh tokens, implement token rotation (issue a new refresh token on each use) and detect reuse to revoke the token family.

Security limits

• PKCE protects the authorization code exchange but does not protect the access token from XSS in the browser — use BFF pattern for higher security.
• All token validation happens locally; the client never sends the code verifier to any party except the authorization server.
• Access tokens in localStorage are vulnerable to XSS — prefer in-memory storage with silent refresh via refresh token.