I was thinking about how to use the homeomorphic property of that cryptosystem to achieve not a "semantic security" but a less ambitous obfuscation of the source code (an ipotetical scenario in which a third party execute the obfuscated code, and get a still obfuscated result). So i was asking myself how much computation time to break a small keysize? i am sure there is a way to calculate "resistance" per bit but i don't know how.
As far as we know, the most efficient attack against well-implemented Paillier encryption is to factor the public modulus $n$ part of the public key. That's also the best known attack against RSA, thus is well-studied.
A smaller data point: CADO-NFS factors a 337 bit $n$ in 5 mn on a single 32-core system.
How much does it cost to break a 64bit keypair? And a 128bit? And so on till 256 or so?
Nothing worth consideration if there is a single such problem to solve.