ConvertX vulnerabilities

In ConvertX before 0.19.0, converters/calibre.ts does not block recipe files and passes them to the ebook-convert program from Calibre. This allows executable code in .recipe or .downloaded_recipe files.

In ConvertX before 0.18.0, the /delete/{id} endpoint uses HTTP GET for permanent deletion without CSRF protection. If an authenticated user visits attacker-controlled content, the browser sends the session cookie and deletes the targeted job.

In ConvertX before 0.18.0, the ImageMagick converter uses the unsanitized convert_to parameter to construct an output path. An authenticated user (or unauthenticated when access is allowed) can upload an extensionless image and supply directory traversal sequences in convert_to via /convert, writing output outside the intended directory.

ConvertX through 0.19.0 contains an arbitrary file read vulnerability that allows authenticated users to read server files because src/converters/pandoc.ts invokes Pandoc without the --sandbox flag. Attackers can upload a reStructuredText document with an include directive naming an absolute path, convert it, and download output containing the referenced file's contents.

ConvertX 0.17.0 contains an arbitrary file read vulnerability in the xelatex converter that allows authenticated users to read files by uploading LaTeX files with input directives. Attackers can upload .tex files containing \\input{path} or \\verbatiminput{path} directives to have the TeX engine read arbitrary files accessible to the server process and include them in downloadable PDF output.

ConvertX, a self-hosted online file converter, in versions prior to 0.16.0, the `/upload` endpoint allows an authenticated user to write arbitrary files on the system, overwriting binaries and allowing code execution. The upload function takes `file.name` directly from user supplied data without sanitization, allowing arbitrary file write.