Issue Bug Resolved

#107895missing secure attribute on fe_typo_user session cookie

Simon Child 10/28/25, 21:33 updated 09/15/26, 06:55

Description

I experienced fail of compliance on AVS scan for PCI DSS, due to missing secure attribute on fe_typo_user session cookie. The context was a little unusual.

Vulnerability Details: Session Cookie Does Not Contain the "Secure" Attribute
Compliance Status:Fail

HTTP Cookie missing Secure attribute on port 443.
Set-Cookie: fe_typo_user=deleted; expires=Thu, 24 Oct 2024 13:26:13 GMT; Max-Age=0; path=/; domain=mytypo3site.org; httponly; samesite=lax
GET / HTTP/1.1
Host: mytypo3site.org
Connection: Keep-Alive

The cookie was flagged for deletion (expiry date in the past) and the content was simply ‘deleted’.

The scanning service did not have any login credentials, so this must have been an anonymous session cookie that had expired (the scan ran for nearly 24 hours).

Usually it would not matter because the browser would delete such a cookie. But (perhaps by design, so that they could inspect such a cookie?) the scanning service ‘browser’ did not delete this cookie and the scan assessment complained about the lack of secure attribute on a cookie served on port 443.

  • Is this a bug, that the deletion process also removes the secure attribute?
  • Is it by design, in the belief that it is no longer necessary as the cookie no longer contains any sensitive data? If so, it should be noted that this can cause fail on scanning for PCI DSS compliance.

As a work around, to get through the compliance testing, I edited symfony/http-foundation/Cookie.php such that every cookie always has the secure attribute (my site is entirely https so this does not cause any problem). But perhaps this issue would be better addressed in a more subtle manner!

Reviews 5

History 1116

  1. Oliver Hader commented <span data-processor="relativetime" data-value="2025-10-29T06:01:19+00:00" > 10/29/25, 06:01 </span>

    For the time being, I marked this issue “private” and will move it to the internal security tracker for further investigation.

    Please report potential security related topics via mail to security@typo3.org. Do not report security findings to public issue trackers, social media or similar.

  2. Oliver Hader commented <span data-processor="relativetime" data-value="2025-10-29T06:35:10+00:00" > 10/29/25, 06:35 </span>

    HTTP sequences

    1: POST login form

    Request
    POST /admin/login?loginProvider=1433416747 HTTP/1.1
    Host: ip14.anyhost.it
    Cookie: __Secure-typo3nonce_uAs3XRtOemrGJbJlF660Rw=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJub25jZSI6InVVZ3JTY2twMnlBNTRMMC1kdHBQZFI5d2N3QmVGcjdyM0YzWjhsMmZaQ2xZRFdjbl9YT0IyQSIsInRpbWUiOiIyMDI1LTEwLTI5VDA3OjIxOjEyKzAxOjAwIn0.fb-sanR1x-agevRbwxqN2o0u8s8hJEIUny2xSNETLfc
    Response
    HTTP/1.1 303 See Other
    Set-Cookie: be_typo_user=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpZGVudGlmaWVyIjoiYWVkNzQzOGJjNGM3NmIxMTZkN2RkNGUwMTI4YTQ4MzAiLCJ0aW1lIjoiMjAyNS0xMC0yOVQwNzoyMjoyMSswMTowMCIsInNjb3BlIjp7ImRvbWFpbiI6ImlwMTQuYW55aG9zdC5pdCIsImhvc3RPbmx5Ijp0cnVlLCJwYXRoIjoiLyJ9fQ.HQKueejwNI9vOOKxhVHg7P_2vS0cIDDpxMai8Dd1IKc; path=/; secure; httponly; samesite=strict
    Set-Cookie: __Secure-typo3nonce_uAs3XRtOemrGJbJlF660Rw=deleted; expires=Tue, 29 Oct 2024 06:22:20 GMT; Max-Age=0; path=/; secure; httponly; samesite=strict
    location: /admin/main?token=f733ad70ec10a09e1bea94e3b1fd30c0fdb188dd

    2: GET logout action

    Request (not relevant, just for completeness)
    GET /admin/logout?token=c9af52fcd5ecc8daa459388a1c75f32c4679f4fb HTTP/1.1
    Host: ip14.anyhost.it
    Cookie: be_typo_user=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpZGVudGlmaWVyIjoiYWVkNzQzOGJjNGM3NmIxMTZkN2RkNGUwMTI4YTQ4MzAiLCJ0aW1lIjoiMjAyNS0xMC0yOVQwNzoyMjoyMSswMTowMCIsInNjb3BlIjp7ImRvbWFpbiI6ImlwMTQuYW55aG9zdC5pdCIsImhvc3RPbmx5Ijp0cnVlLCJwYXRoIjoiLyJ9fQ.HQKueejwNI9vOOKxhVHg7P_2vS0cIDDpxMai8Dd1IKc
    Response ℹ️
    HTTP/1.1 303 See Other
    Set-Cookie: be_typo_user=deleted; expires=Tue, 29 Oct 2024 06:22:29 GMT; Max-Age=0; path=/; domain=ip14.anyhost.it; httponly; samesite=lax
    location: https://ip14.anyhost.it/admin/login

    → sending the broken Set-Cookie: be_typo_user=deleted cookie the first time

    3: GET login form

    Request ⚠️
    GET /admin/login HTTP/1.1
    Host: ip14.anyhost.it
    Cookie: be_typo_user=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpZGVudGlmaWVyIjoiYWVkNzQzOGJjNGM3NmIxMTZkN2RkNGUwMTI4YTQ4MzAiLCJ0aW1lIjoiMjAyNS0xMC0yOVQwNzoyMjoyMSswMTowMCIsInNjb3BlIjp7ImRvbWFpbiI6ImlwMTQuYW55aG9zdC5pdCIsImhvc3RPbmx5Ijp0cnVlLCJwYXRoIjoiLyJ9fQ.HQKueejwNI9vOOKxhVHg7P_2vS0cIDDpxMai8Dd1IKc

    → previous session identifier is still submitted (means: the cookie deletion failed)

    Response ℹ️
    HTTP/1.1 200 OK
    Set-Cookie: be_typo_user=deleted; expires=Tue, 29 Oct 2024 06:22:29 GMT; Max-Age=0; path=/; domain=ip14.anyhost.it; httponly; samesite=lax
    Set-Cookie: __Secure-typo3nonce_ADDsxoJOAXdimUvhJ5phqQ=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJub25jZSI6IjBCN1gtVlBuNGZ0VGYxUENYNWViVDE5VVQ1ZUM4Qk53ZGNKQk1CeUVheEh6OHlpTDZHQnV4USIsInRpbWUiOiIyMDI1LTEwLTI5VDA3OjIyOjMwKzAxOjAwIn0.ZCGicElIwuCGaF12n9YVy-iMbBxqy6YN1rMI0cS14Ko; path=/; secure; httponly; samesite=strict

    → requested session identifier does not exist, TYPO3 performs logoff again, sending the broken Set-Cookie: be_typo_user=deleted cookie again

  3. Oliver Hader commented <span data-processor="relativetime" data-value="2025-10-29T06:42:00+00:00" > 10/29/25, 06:42 </span>

    Every cookie attribute that was used when setting the cookie needs to be used again to override (“delete”) the cookie. This includes domain, httponly and secure (using the same samesite value is not required for the deletion).

    It would be possible to send two Set-Cookie deletion headers (one with secure attribute, one without).

    Side-note: Since we’re using JWTs, when can integrate those attributes that are required for deleting cookie later into the JWT as well at creation time.

  4. Oliver Hader commented <span data-processor="relativetime" data-value="2025-10-29T07:20:04+00:00" > 10/29/25, 07:20 </span>

    Since the session record was removed on the server-side, I cannot see a confidentiality or integrity impact here.
    Re-sending the previous session identifier might be used for tracking (which might be a data privacy issue, but not a confidentiality issue).

    Besides that there is no increased resource consumption or logging when using a session identifier, that does not exist on the server anymore (no availability impact).

    From my point of view, this issue could be handled in public as regular bugfix (for PCI DSS compliance).

  5. Oliver Hader commented and changed project id status Needs Feedback Accepted <span data-processor="relativetime" data-value="2025-10-30T07:20:54+00:00" > 10/30/25, 07:20 </span>

    Conclusion: No security handling & regular public bugfix

  6. Gerrit Code Review pushed patch set 3 for main to change 95713 status Accepted Under Review <span data-processor="relativetime" data-value="2026-09-09T11:43:01+00:00" > 09/09/26, 11:43 </span>

  7. Benjamin Franzke commented and changed status Under Review Resolved done 0 % 100 % <span data-processor="relativetime" data-value="2026-09-15T06:55:10+00:00" > 09/15/26, 06:55 </span>

    Applied in changeset commit:62a11d46011b866b6d5198f433b5672e4fa9df33.