Filevane PDF Upload Audit — dataset
===================================

Version:      2026-08-11
Licence:      CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)
Attribution:  Filevane PDF Upload Audit — https://filevane.com/pdf-upload-audit/
Method:       https://filevane.com/privacy-proof/#automated
Harness:      scripts/upload-audit.mjs

WHAT WAS MEASURED
-----------------
An identical 3,145,728-byte (3.00 MB) PDF, padded with incompressible random
data, was handed to each tool's compress-PDF page. A headless Chromium was
routed through a local counting proxy, and bytes were attributed per destination
host at the transport layer, starting from the instant the file was selected.

The padding matters: a compressible test file could be shrunk before transmission
and read as a partial upload. Encryption is not an obstacle to this method —
TLS hides a request's contents, never its volume.

VERDICT RULES
-------------
uploaded         An observed transfer of at least half the test file to a named
                 host. An observation cannot be un-observed: a tool measured
                 uploading on any run keeps this verdict even if a later run
                 fails to drive it.
local            BOTH near-zero outbound bytes AND independent confirmation that
                 the tool actually ran. Silence alone is not evidence — a tool
                 that failed to start looks identical to one that processed in
                 the browser. Two tools in this set initially read as local for
                 exactly that reason.
not-established  No valid measurement was obtained. THIS IS NOT A FINDING ABOUT
                 THE TOOL and must not be reported as one.

RESULTS IN THIS VERSION
-----------------------
Total tools tested:  20
Uploaded:            12
Local:               1   (Filevane, the control)
Not established:     7

REPRODUCING
-----------
Re-running the harness reproduces this table. If a tool changes its
architecture, this table is wrong until someone re-runs it — which is why every
row carries its own measurement date.

COLUMNS
-------
tool               Name of the tool as published by its operator
url                The exact page tested
date_tested        ISO date of this measurement
verdict            uploaded | local | not-established
bytes_sent         Bytes transmitted to the top destination after the file was selected
destination_host   Where the bytes went; empty when nothing was sent or nothing established
processing_model   server-side | client-side | not-established (derived from verdict)
evidence           Byte count against the 3145728-byte test file, or the reason no measurement was obtained
is_control         TRUE for Filevane, which is the self-test run by the identical method