Perhaps. But it's hard to say. Let me construct a scenario, and tell me how you (or anyone!) would notice:
Some ciphers work on blocks of fixed size, and add padding to reach this length if message is shorter. (example: message must be n*16 bytes, if not, pad message with random bytes at the end, until it is.)
Let's say I've backdored a program implementing such a cipher. The backdoor is this: Instead of padding with random bytes, I do this:
1) Take as much of the secret key as will fit in the padding-space. (if 9 bytes of padding is needed, I take the first 9 bytes of the secret key)
2) I encrypt this (using a algorithm that can encrypt any-length messages) using a second hidden backdoor-key.
3) I swap the last n bytes of the ciphertext with this encrypted partial-key.
Result: Message-size is unchanged. Encryption and Decryption works as specified. n-last characters (the padding) looks like random noise, and is supposed to BE random. How do you notice ? How do you detect that the last n characters is really part of the key, encrypted, and NOT random noise ?
(To make this more fun: I left one big flaw in the scheme there IS a easy way to detect that this shit is going on -- but there's also a way to patch that flaw, I'll explain that in the next message if you find the flaw)