Invalid crypto padding

The word before "padding" names the segment: `header`, `payload` or `crypto` (the signature).

PyJWT (Python) — jwt.exceptions.DecodeError while base64url-decoding a segment

What it means

PyJWT split the token into three segments but could not base64url-decode one of them. "crypto padding" refers to the third (signature) segment, "header padding" and "payload padding" to the first and second. JWS segments must use the URL-safe base64 alphabet (`A–Z a–z 0–9 - _`) without `=` padding; spaces, `+`, `/` or a wrong length make decoding fail before any signature check.

Why it happens

• The Bearer prefix is still attached

  common

  Passing `"Bearer eyJ…"` puts a space and the word Bearer into the header segment, which then fails with "Invalid header padding".

• A producer used standard base64

  common

  Hand-written signers that use `base64.b64encode` instead of the URL-safe variant, or that keep the `=` padding, produce segments with `+`, `/` or `=`. Most often this hits the signature segment, giving "Invalid crypto padding".

• The token was URL-encoded or decoded on the way

  occasional

  Tokens passed in query strings or form fields can arrive with `%2B`/`%3D` sequences or with `+` turned into a space by a form decoder.

• Characters were added or lost in a copy

  occasional

  Line breaks from a terminal, quotes from JSON, or an ellipsis inserted by a log viewer change the segment lengths and alphabet.

How to fix it

1. 1.

   Strip the scheme and validate the alphabet

   Accept only `Bearer <token>` where the token matches the base64url JWS pattern; answer 401 for everything else before calling PyJWT.

   import re
   
   BASE64URL_JWS = re.compile(r"^[A-Za-z0-9_-]+\.[A-Za-z0-9_-]+\.[A-Za-z0-9_-]*$")
   
   def extract(authorization: str) -> str:
       scheme, _, token = authorization.partition(" ")
       token = token.strip()
       if scheme.lower() != "bearer" or not BASE64URL_JWS.match(token):
           raise ValueError("not a compact JWS")
       return token

2. 2.

   Encode with base64url and no padding

   If you build tokens yourself, use URL-safe base64 and remove trailing `=`. Better: let a JOSE library assemble the token.

   import base64
   
   def b64url(data: bytes) -> str:
       # JWS uses base64url WITHOUT padding (RFC 7515 section 2).
       return base64.urlsafe_b64encode(data).rstrip(b"=").decode("ascii")

← Browse the full signature error database