[stuaxo]

An s3 bucket will give you back an ETAG header for each file it holds, which is more or less the MD5 of a set of chunks that make up the file.

You can use that to identify files.

[orf]

Only if the file is not uploaded via a multipart upload, which all SDKs do for any files larger than a megabyte.

If the object is encrypted with KMS, which almost everything is, then the hash is the encrypted contents and not the original contents.