I use AES-CTR to perform on the fly encryption of packets. Each packet is 16 bytes long, however, depending on what is being transferred, the 16 bytes may not be used. So far, unused bytes are left to 0.
I'm not a cryptographer, but I'm wondering how helpful this can be to an attacker to have such an information. Knowing the protocol, you can guess a few bytes per packet in general, but with these bytes left to 0, you can actually guess most bytes.
Would it be recommended to instead use random padding instead of zeroes ?