Offhand, it seems to me that
symmetric encryption would be perfectly satisfactory for your situation. (Either GPG or
openssl could be used, because all of them provide good, trustworthy implementations of good, trustworthy symmetric ciphers.)
You simply want to encrypt the data using a
(symmetric) key that you believe that you can "keep secret." You simply want to be sure that anyone who steals the backup cannot make use of it without simultaneous possession of the key.
You are not
transferring these files anywhere, and this
(IMHO ...) negates any practical necessity to try to use "public key" (PKI) techniques.
In all PKI situations, some (pure-random)
symmetric cipher-key is actually used
(due to practical necessity ...) to encrypt the data, while the PKI layer is used to enable the two parties to transfer knowledge of that symmetric key's value (in this particular message). You have no need for this layer of complexity, and it brings no benefit whatsoever to you.
Therefore, keep it simple: symmetrically encrypt the data using a random,
secret, binary, key. Make damned-sure that no one can penetrate the software to extract the key. Protect access to the key by appropriate
conventional means, while being sure that no one will resort to writing it down on a piece of tape underneath the keyboard, just "to get their
work done."
Be sure that the key, far from being
anything that "could possibly be derived from 'anything that might be found in a dictionary,'" is an absolutely-unpredictable
random number which is the same size as the underlying symmetric key-space. Then,
guard it.
Incidentally, the
openssl tool has some very useful-to-you functions, such as
openssl rand, which "generates pseudo-random bytes." That would be a terrific way to generate your symmetric key, because this is a
cryptographically strong PRNG = Pseudo-Random Number Generator. (Far superior, then, to "
rand()," or to "haphazard banging-around on your keyboard.") This trustworthy tool
really can produce "an
n-bit key" that actually
would give your opponent "2^
n possibilities, every one of them
equally likely." Which is exactly what you want for "a symmetric cipher-key" that you think that you can actually manage to "keep secret."