# how many pairs of plaintext-ciphertext are needed before we break AES? [duplicate]

Imagine a set of words ($plaintext$) each are encrypted using AES-ECB mode and due to some leakage in the system an attacker could gain access to some pairs of $plaintext$ and $ciphertext$. But once the attacker got bunch of such pairs he cannot further ask the oracle for any other pairs. So in reality how many such pairs of $plaintext$ and $ciphertext$ are needed to crack the $key$.

Here each input $|plaintext| < 128$ bits.

• As far as I know, you can have all the plaintext/ciphertext pairs (there is a finite number, thanks to the restriction in the question) and still not be able to extract the AES key. – mikeazo Aug 17 '15 at 14:46
• @mikeazo ...in reasonable time, that is. The question did not make any restriction on the running time of an attack (but admittedly does not make too much sense without that implicit addition). – yyyyyyy Aug 17 '15 at 14:58
• @yyyyyyy, yeah, that is what I was thinking too. Also, with reason amount of memory. – mikeazo Aug 17 '15 at 15:01
• @yyyyyyy I think the expected life time of our Sun is reasonable enough. Otherwise: the expected life time of the universe? – Maarten Bodewes Aug 17 '15 at 15:20
• so does this mean , we cannot infer the key in any reasonable amount of time ? – sashank Aug 17 '15 at 15:21