Please be cautious: EncFS was not designed to resist an attacker having ongoing access to the volume, as they would in this scenario!

In particular, an attacker having access to the ciphertext at two or more different times violates EncFS's security assumptions; undetected malicious modification of files is also feasible in this scenario.

Many encrypted filesystems do not include such a property in their design criteria - for example, XTS mode as-is is not suitable for use in this scenario either, so please also try to avoid putting TrueCrypt (et al) on Dropbox!

For a broad general example of what a system would look like which tries to address this use case more naturally and effectively (although, caveat: I have not reviewed it in great detail myself), please see Tahoe-LAFS.