CVE-2026-101051

Cloudreve before 4.16.1 fails to properly sanitize file paths returned by remote downloaders, allowing authenticated users to create files outside the selected destination directory. Attackers can exploit path traversal sequences in downloader metadata to write files to unexpected locations within accessible namespaces.

Cloudreve before 4.16.1 fails to properly sanitize file paths returned by remote downloaders, allowing authenticated users to create files outside the selected destination directory. Attackers can exploit path traversal sequences in downloader metadata to write files to unexpected locations within accessible namespaces.

Summary

Cloudreve trusts file paths returned by the configured remote downloader. A downloader-reported path such as ../../escaped.txt can cause a downloaded file to be created outside the user-selected destination directory.

Details

In the remote download master transfer path, Cloudreve joins the user-selected destination URI with the downloader-reported file name.

// pkg/filemanager/workflows/remote_download.go:436-438
sanitizedName := sanitizeFileName(file.Name)
dst := dstUri.JoinRaw(sanitizedName)
src := filepath.FromSlash(path.Join(m.state.Status.SavePath, file.Name))

The same issue also exists when constructing slave upload payloads.

// pkg/filemanager/workflows/remote_download.go:323-327
dst := dstUri.JoinRaw(sanitizeFileName(f.Name))
src := path.Join(m.state.Status.SavePath, f.Name)
payload.Files = append(payload.Files, SlaveUploadEntity{
	Src:   src,
	Uri:   dst,

The sanitizer does not remove /, ., or .. path segments.

// pkg/filemanager/workflows/remote_download.go:648-650
func sanitizeFileName(name string) string {
	r := strings.NewReplacer("\\", "_", ":", "_", "*", "_", "?", "_", "\"", "_", "<", "_", ">", "_", "|", "_")
	return r.Replace(name)
}

JoinRaw() splits the raw string by / and joins the segments, allowing .. to affect the final URI path.

// pkg/filemanager/fs/uri.go:173-175
func (u *URI) JoinRaw(elem string) *URI {
	return u.Join(strings.Split(strings.TrimPrefix(elem, Separator), Separator)...)
}

For aria2, Cloudreve derives downloader.TaskFile.Name from the path returned by aria2.tellStatus().files[].path.

// pkg/downloader/aria2/aria2.go:148-159
relPath := strings.TrimPrefix(filepath.ToSlash(item.Path), savePath)
if len(relPath) > 0 {
	relPath = relPath[1:]
}
return downloader.TaskFile{
	Index:    index,
	Name:     relPath,

Therefore, if the selected destination is: cloudreve://my/victim/safe, the downloader reports ../../escaped.txt, the final upload destination becomes cloudreve://my/escaped.txt

The issue can move the final Cloudreve URI further up the user’s any accessible namespace, but is subject to Cloudreve’s normal permission and upload checks.

PoC

The PoC uses a fake aria2 JSON-RPC service to simulate a downloader returning a traversal path. The vulnerable input is downloader metadata returned by the downloader API, not the HTTP response body of the downloaded URL.

Setup:

Cloudreve official Docker image
PostgreSQL
Redis
Fake aria2 JSON-RPC service

Configure the master node in the Cloudreve admin UI:

Remote download capability: enabled
Downloader provider: aria2
aria2 RPC server: http://fake-aria2:6800/jsonrpc
aria2 token: empty

Create this folder structure in the file manager:

My files /
  victim /
    safe /

Create a remote download task using any URL in the victim/safe directory, for example:

http://attacker.invalid/file

The fake aria2 service returns:

files[0].path = <saveDir>/../../escaped.txt

Expected result after the remote download task completes:

cloudreve://my/escaped.txt exists

This demonstrates that the downloaded file escapes both the selected destination directory and its parent directory.

Impact

If a configured remote downloader returns malicious file metadata, Cloudreve may create downloaded files outside the destination directory selected by the user who starts the remote download task.

This affects authenticated users who have remote-download permission and create remote download tasks. The resulting file is still subject to Cloudreve’s normal upload and permission checks, but it may be placed in an unexpected writable location outside the selected folder.

Appendix: fake_aria2.py

import json
import os
from http.server import BaseHTTPRequestHandler, HTTPServer

GID = "0123456789abcdef"
CONTENT = b"created outside the selected Cloudreve destination\n"
save_dir =…

Cloudreve before 4.16.1 fails to properly sanitize file paths returned by remote downloaders, allowing authenticated users to create files outside the selected destination directory. Attackers can exploit path traversal sequences in downloader metadata to write files to unexpected locations within accessible namespaces.

Cloudreve before 4.16.1 fails to properly sanitize file paths returned by remote downloaders, allowing authenticated users to create files outside the selected destination directory. Attackers can exploit path traversal sequences in downloader metadata to write files to unexpected locations within accessible namespaces.

Cloudreve before 4.16.1 fails to properly sanitize file paths returned by remote downloaders, allowing authenticated users to create files outside the selected destination directory. Attackers can exploit path traversal sequences in downloader metadata to write files to unexpected locations within accessible namespaces.

Summary

Cloudreve trusts file paths returned by the configured remote downloader. A downloader-reported path such as ../../escaped.txt can cause a downloaded file to be created outside the user-selected destination directory.

Details

In the remote download master transfer path, Cloudreve joins the user-selected destination URI with the downloader-reported file name.

// pkg/filemanager/workflows/remote_download.go:436-438
sanitizedName := sanitizeFileName(file.Name)
dst := dstUri.JoinRaw(sanitizedName)
src := filepath.FromSlash(path.Join(m.state.Status.SavePath, file.Name))

The same issue also exists when constructing slave upload payloads.

// pkg/filemanager/workflows/remote_download.go:323-327
dst := dstUri.JoinRaw(sanitizeFileName(f.Name))
src := path.Join(m.state.Status.SavePath, f.Name)
payload.Files = append(payload.Files, SlaveUploadEntity{
	Src:   src,
	Uri:   dst,

The sanitizer does not remove /, ., or .. path segments.

// pkg/filemanager/workflows/remote_download.go:648-650
func sanitizeFileName(name string) string {
	r := strings.NewReplacer("\\", "_", ":", "_", "*", "_", "?", "_", "\"", "_", "<", "_", ">", "_", "|", "_")
	return r.Replace(name)
}

JoinRaw() splits the raw string by / and joins the segments, allowing .. to affect the final URI path.

// pkg/filemanager/fs/uri.go:173-175
func (u *URI) JoinRaw(elem string) *URI {
	return u.Join(strings.Split(strings.TrimPrefix(elem, Separator), Separator)...)
}

For aria2, Cloudreve derives downloader.TaskFile.Name from the path returned by aria2.tellStatus().files[].path.

// pkg/downloader/aria2/aria2.go:148-159
relPath := strings.TrimPrefix(filepath.ToSlash(item.Path), savePath)
if len(relPath) > 0 {
	relPath = relPath[1:]
}
return downloader.TaskFile{
	Index:    index,
	Name:     relPath,

Therefore, if the selected destination is: cloudreve://my/victim/safe, the downloader reports ../../escaped.txt, the final upload destination becomes cloudreve://my/escaped.txt

The issue can move the final Cloudreve URI further up the user’s any accessible namespace, but is subject to Cloudreve’s normal permission and upload checks.

PoC

The PoC uses a fake aria2 JSON-RPC service to simulate a downloader returning a traversal path. The vulnerable input is downloader metadata returned by the downloader API, not the HTTP response body of the downloaded URL.

Setup:

Cloudreve official Docker image
PostgreSQL
Redis
Fake aria2 JSON-RPC service

Configure the master node in the Cloudreve admin UI:

Remote download capability: enabled
Downloader provider: aria2
aria2 RPC server: http://fake-aria2:6800/jsonrpc
aria2 token: empty

Create this folder structure in the file manager:

My files /
  victim /
    safe /

Create a remote download task using any URL in the victim/safe directory, for example:

http://attacker.invalid/file

The fake aria2 service returns:

files[0].path = <saveDir>/../../escaped.txt

Expected result after the remote download task completes:

cloudreve://my/escaped.txt exists

This demonstrates that the downloaded file escapes both the selected destination directory and its parent directory.

Impact

If a configured remote downloader returns malicious file metadata, Cloudreve may create downloaded files outside the destination directory selected by the user who starts the remote download task.

This affects authenticated users who have remote-download permission and create remote download tasks. The resulting file is still subject to Cloudreve’s normal upload and permission checks, but it may be placed in an unexpected writable location outside the selected folder.

Appendix: fake_aria2.py

import json
import os
from http.server import BaseHTTPRequestHandler, HTTPServer

GID = "0123456789abcdef"
CONTENT = b"created outside the selected Cloudreve destination\n"
save_dir =…

Cloudreve's remote download file paths can escape the selected destination directory in github.com/cloudreve/Cloudreve