Multi-Party Computation with TFHE

Zama TFHE-rs
1

Hello,

First of all - I am a researcher and fairly new to encryption as I want to use it as a tool for developing protocols in my field of business informatics.

So what I currently do is, that I have some multi-party computation setting, where I have for example two organizations that want to order the data on some feature, e.g., timestamps. Each organizations have their data sorted on their timestamps, i.e., for organization A there is t_1, t_2, \dots, t_n and for organization B is t_1', t_2', \dots, t_m'.

Now, I suggest that they both iterate over their timestamps and A encrypts the current timestamp t_i, sends the encrypted value over to organization B, organization B sends the encrypted boolean of the comparison t_i \leq t_j' back and based on that organization A tells organization B which organization has to progress with their current position.

My question is now, that I’m unsure whether reporting the truth value to organization B might break the TFHE encryption since the Security section of the TFHE documentation describes:
“In particular, it is assumed that the results of decrypted computations are not shared by the secret key owner with any third parties, as such an action can lead to leakage of the secret encryption key. If you are designing an application where decryptions must be shared, you will need to craft custom encryption parameters which are chosen in consideration of the IND-CPA^D security model [1].”

Could someone give me some instruction on what to do? Or some references that I can look up where someone did something similar?

Kind regards,
Christian

2

Deleted this post as answer was innacurate

1 Like
3

@Crenner I’ll get back to you checking the last part of the ind cpa d warning

1 Like
4

Hello Christian,

The warning in the documentation has been updated: new parameters are now available to ensure security in the IND-CPA^D model. Although the previous parameters were already secure against practical attacks, the new ones guarantee a security level comparable to the standard IND-CPA model.

Without diving too much into the details, the key idea behind the IND-CPA^D model is to allow decrypted values to be shared with an entity that does not have access to the secret decryption key. State-of-the-art attacks in this model exploited potential computational errors to extract partial information about the secret key. Previously, the probability of a computational failure (pfail) was below 2^-64, making practical attacks extremely difficult to carry out. With version 1.0, we now provide parameters ensuring pfail < 2^-128, rendering such attacks negligible and guaranteeing security within this model.

Given your use case, these new parameters should provide the necessary level of security FHE-wise. I guess now this is more about ensuring the security of your protocol regarding MPC models.

J-B.