A set of key pairs and one hash to secure them

Question

I have a simple problem: I have a set of users' ECDSA key pairs, and say I want to encrypt them with a simple algorithm. I have access to one variable that uniquely identifies the user, so I hash it with a SHA-2 algorithm and have a string of bytes longer than each of the keys.

Is it safe to XOR a key pair, or XORing both of the keys together will make it easier to decode the original value?

What if I hash the entire key chain, does that make it much easier to get the data?

Answer 1

What you are proposing in effect means that you use a not-really-random one-time-pad, which is used twice (i.e. a two-times-pad). This is not secure.

1. Using a single hash to generate a key from a password is a bad idea - especially if the password is short, it is easy to brute-force it (i.e. try lots of passwords).

2. Using the simple XOR cipher to encrypt a ECDSA key pair, one half of which is public anyway, means that half your key is directly derive-able from the encrypted key pair. (This helps with the brute-forcing in step 1.)

3. Using a one-time-pad twice means that the attacker can derive $x_1 \oplus x_2$ (i.e. the XOR of the two private keys). I don't know of any way to come from this to $x_1$ and $x_2$ (given also the public keys $g^{x_1}$ and $g^{x_2}$), but I would not wonder if someone would find a way to do this easier than trying all $x_1$.

Instead, use established algorithms:

• Use an established key derivation function (Bcrypt, Scrypt, PBKDF2) with an appropriate work factor to derive the key from a password.
• Use a secure cipher algorithm (a block or stream cipher) in a secure mode instead of XORing.