oauth（OAuth 安全）

Flow Selection

• Authorization Code + PKCE: use for all clients—web apps, mobile, SPAs
• Client Credentials: service-to-service only—no user context
• Implicit flow: deprecated—don't use; was for SPAs before PKCE existed
• Device Code: for devices without browsers (TVs, CLIs)—user authorizes on separate device

PKCE (Proof Key for Code Exchange)

• Required for public clients (SPAs, mobile), recommended for all
• Generate code_verifier: 43-128 char random string, stored client-side
• Send code_challenge: SHA256 hash of verifier, sent with auth request
• Token exchange includes code_verifier—server verifies against stored challenge
• Prevents authorization code interception—attacker can't use stolen code without verifier

State Parameter

• Always include state in authorization request—prevents CSRF attacks
• Generate random, unguessable value; store in session before redirect
• Verify returned state matches stored value before processing callback
• Can also encode return URL or other context (encrypted or signed)

Redirect URI Security

• Register exact redirect URIs—no wildcards, no open redirects
• Validate redirect_uri on both authorize and token endpoints
• Use HTTPS always—except localhost for development
• Path matching is exact—/callback ≠ /callback/

Tokens

• Access token: short-lived (minutes to hour), used for API access
• Refresh token: longer-lived, used only at token endpoint for new access tokens
• ID token (OIDC): JWT with user identity claims—don't use for API authorization
• Don't send refresh tokens to resource servers—only to authorization server

Scopes

• Request minimum scopes needed—users trust granular requests more
• Scope format varies: openid profile email (OIDC), repo:read (GitHub-style)
• Server may grant fewer scopes than requested—check token response
• openid scope required for OIDC—triggers ID token issuance

OpenID Connect

• OIDC = OAuth 2.0 + identity layer—adds ID token and UserInfo endpoint
• ID token is JWT with sub, iss, aud, exp + profile claims
• Verify ID token signature before trusting claims
• nonce parameter prevents replay attacks—include in auth request, verify in ID token

Security Checklist

• HTTPS everywhere—tokens in URLs must be protected in transit
• Validate iss and aud in tokens—prevents token confusion across services
• Bind authorization code to client—code usable only by requesting client
• Short authorization code lifetime (10 min max)—single use
• Implement token revocation for logout/security events

Common Mistakes

• Using access token as identity proof—use ID token for authentication
• Storing tokens in localStorage—vulnerable to XSS; prefer httpOnly cookies or memory
• Not validating redirect_uri—allows open redirect attacks
• Accepting tokens from URL fragment in backend—fragment never reaches server
• Long-lived access tokens—use short access + refresh pattern

Token Endpoints

• /authorize: user-facing, returns code via redirect
• /token: backend-to-backend, exchanges code for tokens; requires client auth for confidential clients
• /userinfo (OIDC): returns user profile claims; requires access token
• /revoke: invalidates tokens; accepts access or refresh token

Client Types

• Confidential: can store secrets (backend apps)—uses client_secret
• Public: cannot store secrets (SPAs, mobile)—uses PKCE only
• Never embed client_secret in mobile apps or SPAs—it will be extracted