Often times, (asymmetric) encryption schemes can be discussed in terms of provable security - we can show that the scheme is secure under certain attacks if the underlying cryptographic primative is hard. In other words, provable security relies on reducing our problem to its primative, for example, the RSA problem is reducable to the factoring problem.
Generally, the following four properties are discussed, IND, CPA, CCA, and CCA2:
IND (indistinguishibility): An adversary cannot distinguish the encryption of any two messages of the same length ie. if given a challenge cipher text $c$, they cannot tell if $c$ came from $m_1$ or $m_2$.
IND-CPA (indistinguishibility under a chosen plaintext attack): An adversary cannot distinguish which message was used to create a challenge ciphertext if given access to the public key (in other words, they are able to create ciphertexts from chosen messages).
IND-CCA (indistinguishibility under a chosen ciphertext attack): The same set up as before, however this time, the adversary has a decryption oracle they are able to query until they are given the challenge ciphertext.
IND-CCA2 (indistinguishibility under an adaptive chosen ciphertext attack): Like the previous example, however now the adversary is allowed to query the decryption oracle even after recieving the challenge (with the caveat they are not allowed to enter the challenge ciphertext to the oracle)
These are chosen to be discussed as if we can prove that (assuming the proof applied in the proper context), we only need worry about the hardness of the problem underlying it.
The definitions of these properties may be slightly different for symmetric schemes (they are not based on mathematical problems), but we are able to prove analogous results, again under certain conditions (think key length).
This allows us to now assess the 'security level' for these parameters in a way that can be translated and compared to that of AES.