CVE-2026-55375: canto-saas-api: OAuth credentials logged via query params

## Summary In affected versions, the OAuth2 token request sends `app_id`, `app_secret`, `refresh_token` and `code` as URL query parameters of the POST request to `https://oauth.<domain>/oauth/api/oauth2/token`. Request URLs are commonly recorded in access logs, proxy logs and APM traces, so the application secret and refresh token can be persisted in plain text outside the application's control. In addition, when the token request fails, the Guzzle exception message — which contains the full request URI including the credentials — was passed unmodified into the `AuthorizationFailedException` thrown by `OAuth2::obtainAccessToken()`. Applications that log exceptions or forward them to error trackers (e.g. Sentry) may therefore have recorded the app secret in their logs. ## Impact An attacker with access to web server logs, proxy logs, APM tracing data or application error logs of a consumer of this library can obtain the Canto `app_secret`, `refresh_token` or authorization `code` and use them to obtain access tokens for the Canto tenant. ## Patches Fixed in 3.0.0: - OAuth credentials are sent in the form-encoded POST body instead of the URL query string (RFC 6749 §2.3.1). `OAuth2Request::getQueryParams()` now returns `null`; the parameters are available via `getFormParams()`. - Exception messages are sanitized before being rethrown: the values of `app_secret`, `refresh_token` and `code` are masked (including url-encoded, differently cased and JSON-embedded variants). ## Workarounds If you cannot upgrade: - Treat web server, proxy and APM logs of systems performing Canto OAuth requests as secret material and restrict access to them. - Catch `AuthorizationFailedException` in your application and strip the query string from the message before logging or forwarding it. If your logs may have been exposed, rotate the affected Canto app secret.

An attacker with access to a victim organization's centralized log platform — via a compromised DevOps credential, a misconfigured Elasticsearch/OpenSearch cluster, or through a third-party APM SaaS (Datadog, New Relic) with overly broad data access — queries for HTTP traffic to the Canto OAuth token endpoint. Because the library passes app_secret and refresh_token as URL query parameters, they appear in the request URL recorded by every intermediate logging layer. The attacker extracts the credentials and submits a token request to Canto, obtaining a valid access token. With this token, they authenticate as the victim application, browse and exfiltrate the organization's Canto tenant (including AI training data, proprietary media, or labeled datasets), and rotate the credentials to maintain persistence — all while the legitimate application continues operating, masking the intrusion.