Re:Encryption is dead, and we killed it
by viraptor on Sunday January 08, 2017 @11:22PM
(#13353984)
Ok, so I'm going to only comment on some parts.
First: They're proving that they can generate the encryption key based on password and salt. Which is exactly how you generate the encryption key. "We have the full key. We only used MD5 and didn't write a single line of code." - regardless of whether the key is generated in the described way, everything's fine. This is exactly the input and output elements you'd expect.
Regarding key generation / KDF, you can see it in the official docs as well: https://www.openssl.org/docs/man1.0.1/crypto/EVP_BytesToKey.... which actually say you should probably go with PKCS#5 2.0 instead. But really with high enough count, it shouldn't matter that much if the hash is MD5. Collisions have been found, but it's not collisions you'd be looking for when decrypting messages. You'd need to find the original string from multiple applications of MD5, which does effectively involve bruteforce.
The last part: "The question is, what the hell does this number mean, and how do we interpret these results?" - this is actually something you can find in the source. No need to raise it as an interesting question. (for the curious, it's the CRYPTO_THREAD_ID https://github.com/openssl/openssl/blob/6f0ac0e2f27d9240516e... )
"This number is, also, what the padding oracle is mysteriously spitting out." -> it's actually what your memory fragmentation / aslr is likely spitting out. The number is affected only by the system, not by the contents of the file.
First: They're proving that they can generate the encryption key based on password and salt. Which is exactly how you generate the encryption key. "We have the full key. We only used MD5 and didn't write a single line of code." - regardless of whether the key is generated in the described way, everything's fine. This is exactly the input and output elements you'd expect.
Regarding key generation / KDF, you can see it in the official docs as well: https://www.openssl.org/docs/man1.0.1/crypto/EVP_BytesToKey.... which actually say you should probably go with PKCS#5 2.0 instead. But really with high enough count, it shouldn't matter that much if the hash is MD5. Collisions have been found, but it's not collisions you'd be looking for when decrypting messages. You'd need to find the original string from multiple applications of MD5, which does effectively involve bruteforce.
The last part: "The question is, what the hell does this number mean, and how do we interpret these results?" - this is actually something you can find in the source. No need to raise it as an interesting question. (for the curious, it's the CRYPTO_THREAD_ID https://github.com/openssl/openssl/blob/6f0ac0e2f27d9240516e... )
"This number is, also, what the padding oracle is mysteriously spitting out." -> it's actually what your memory fragmentation / aslr is likely spitting out. The number is affected only by the system, not by the contents of the file.
This is just trolling. It happens.