Encode via FixedPointEncoder. Args: data (torch.Tensor): Tensor to be encoded Returns: encoded_data (torch.T

Computes shares of random number in Zp*.Zp* is the multiplicative group mod p. Args: session (Session): session to generate rando

Prepare the share tensors by calling the reshape function in parallel at each party. Args: x (MPCTensor): the MPCTensor on which to apply

PRZS generator. Setup the Pseudo-Random-Zero-Share generators to the parties involved in the communication. Assume there are

Compute the tanh using the sigmoid approximation. Args: tensor (tensor): values where tanh should be approximated Returns: t

Add SyMPC methods (only "share") to the torch.Tensor. Raises: ValueError: if torch.Tensor has already the method added.

Concatenates the given sequence of seq tensors in the given dimension. Args: tensors (List): sequence of tensors to concatenate d

r"""Computes Chebyshev coefficients. For n = terms, the ith Chebyshev series coefficient is c_i = 2/n \sum_{k=1}^n \cos(j(2k-1)\pi / 4n)

Apply the "conv_transpose2d" operation between "self" and "y". Args: weight: the convolution kernel bias: optional bi

Copy attributes from torch layer to SyMPC layer. Args: layer (torch.nn.modules): The torch layer layer_name (str): layer name

Distribute a list of shares. Args: shares (List[ShareTensor): list of shares to distribute. session (Session): Sessio

Forward the call to the tensor specific class. Args: *args (List[Any]): list of args to be forwarded **kwargs(Dict[st

Distributes shares to party. Args: shares (List[Union[ShareTensor,torch.Tensor]]): Shares to be distributed. party_ra

Equal operator. Args: other (MPCTensor): MPCTensor to compare. Returns: MPCTensor: Result of the comparison.

Compare two SMPC modules. Args: layer1 (SMPCModule): the first operand for the comparison layer2 (SMPCModule): the se

Perform the feedforward and compute the result for the transpose operation. Args: ctx (Dict[str, Any]): Context used to save info

Generate a list of clients given a number. Returns: Callable[[int], List[Any]]: List of clients

Get the state of a CryptoProvider. Returns: str: CryptoProvider

Get a torch module from a given MPC Conv2d module. The parameters of the models are not set. Args: conv_module (Conv2d):

Greater than operator. Args: other (MPCTensor): MPCTensor to compare. Returns: MPCTensor: Result of the comp

Calculates hardtanh of given tensor. Defined as 1 if x > 1 -1 if x < -1 x otherwise Args: tensor (MPCTensor)

Initialize the generators - that are used for Pseudo Random Zero Shares. Args: seed_current (int): the seed for our party

Apply the "isub" operation between "self" and "y". Args: y (Union["MPCTensor", torch.Tensor, float, int]): self - y Retu

Performs local decomposition to generate shares of shares. Args: x (ReplicatedSharedTensor) : input RSTensor. ring_si

Apply the "matmul" operation between "self" and "y". Args: y (Union[int, float, torch.Tensor, "ShareTensor"]): self @ y.

Apply the "matmul" operation between "self" and "y". Args: y (Union["MPCTensor", torch.Tensor, float, int]): self @ y Re

Mean Squared Error loss. Args: pred (MPCTensor): The predictions obtained target (MPCTensor): The target values reduction

Falcon malicious multiplication. Args: x (MPCTensor): Secret y (MPCTensor): Another secret session (Sessi

Reconstruct the shared state dict. Returns: Dict[str, Any]: The reconstructed state dict.

Sanity check to validate that a new instance for MPCTensor can be created. Args: secret (Union[ShareTensor, torch.Tensor, float,

Create a convolutional network and do a feedforwad. Args: get_clients: Fixture that returns a list of clients Returns: A tup

Share the parameters of the normal Conv2d layer. Args: state_dict (Dict[str, Any]): the state dict that would be shared.

Sigmoid function. Args: x (MPCTensor): The tensor on which we apply the function Returns: An MPCTensor which represents the

Checks if operation requires truncation and performs it if required. Args: input_tensor (MPCTensor): Result of operation

Equal operator. Check if "self" is equal with another object given a set of attributes to compare. Args: oth

Check if "self" is equal with another object given a set of attributes to compare. Args: other (Any): Object to compare

Check if "self" is equal with another object given a set of attributes to compare. Args: other (Any): Object to compare

Check if "self" is equal with another object given a set of attributes to compare. Args: other (Any): Object to compare

Check if "self" is equal with another object given a set of attributes to compare. Args: other (Any): Session to compare.

Get the attribute and check if we should track the gradient. Args: attr_name (str): The name of the attribute Returns:

Allows to subset shares. Args: key (int): The share to be retrieved. Returns: share (torch.Tensor): Returned

Greater than operator. Args: y (Union["ShareTensor", torch.Tensor, int]): Tensor to compare. Returns: bool:

The initializer for the Linear layer. Args: session (Session): the session used to identify the layer

Initialize Conv2d layer. The stride, padding, dilation and groups are hardcoded for the moment. Args: session (Session):

Initialize ReplicatedSharedTensor. Args: shares (Optional[List[Union[float, int, torch.Tensor]]]): Shares list fr

Initialize ShareTensor. Args: data (Optional[Any]): The share a party holds. Defaults to None config (Config): The co

Initializer for the MPCTensor. It can be used in two ways. ShareTensorControlCenter can be used in two ways: - secret is known by the

Initialization of the Protocol. Args: security_type : specifies the security level of the Protocol. Raises:

Initialization of the Protocol. Args: security_type : specifies the security level of the Protocol. Raises:

Initialization of the Protocol. Args: security_type : specifies the security level of the Protocol. Raises:

Initialize a new SGD optimizer. Args: parameters (List[MPCTensor]): Parameters of the module to optimize lr (float):

Initialize FP Encoder. Args: base (int): The base for the encoder. precision (int): The precision for the encoder.

Initializer.

Initializer for the Session Manager. Raises: NotImplementedError: This class it is not supposed to be instanciated.

Initializer for the Session. Args: parties (Optional[List[Any]): Used to send/receive messages: ring_size (int): Fiel

Session configuration. Config can be used inside a Session (see Session) to specify different options that can be used for the Fixed

Initialize convolutional network. Arguments: torch_ref: Reference to the torch library

Return the length of MPCTensor. Returns: int: Length

Lower than operator. Args: y (Union["ShareTensor", torch.Tensor, int]): Tensor to compare. Returns: bool: Re

Add new methods and properties to the defined tensor classes. Args: name (str): name of the class which has SyMPCTensor metaclass

Control creation of new instances. Args: name (str): Name of the protocol bases: asdf. dct (Dict[Any, Any

Representation. Returns: str: Representation.

Allows to set share value to new value. Args: key (int): The share to be retrieved. newvalue (torch.Tensor): New valu

Representation. Returns: str: The representation.

Generates truncation pair for the given MPCTensor. Args: x (MPCTensor): input tensor session (Session) : session of t

Function used to request and get a share - Duet Setup. Args: share_ptr (ReplicatedSharedTensor): input ReplicatedSharedTensor

Function used to request and get a share - Duet Setup. Args: share_ptr (ShareTensor): a ShareTensor Returns:

Apply the "add" operation between "self" and "y". Args: y (Union[int, float, torch.Tensor, "ReplicatedSharedTensor"]): self + y

Apply the "add" operation between "self" and "y". Args: y (Union[int, float, torch.Tensor, "ShareTensor"]): self + y Ret

Apply the "add" operation between "self" and "y". Args: y (Union["MPCTensor", torch.Tensor, float, int]): self + y Retur

Computes addition(x+y) modulo PRIME_NUMBER constant. Args: x (torch.Tensor): input tensor y (torch.tensor): input ten

Compute argmax using pairwise comparisons. Makes the number of rounds fixed, here it is 2. This is inspired from CrypTen. Args: x (M

Perform the backward pass and compute the gradient for the implemented operation. Args: ctx (Dict[str, Any]): Context used to sav

Perform the backward pass for the addition operation. Args: ctx (Dict[str, Any]): Context used to retrieve the information for th

Perform the backward pass for the summation operation. Args: ctx (Dict[str, Any]): Context used to retrieve the information for t

Perform the backward pass for the substraction operation. Args: ctx (Dict[str, Any]): Context used to retrieve the information fo

Perform the backward pass for the substraction operation. Args: ctx (Dict[str, Any]): Context used to retrieve the information f

Perform the backward pass for the multiplication operation. Args: ctx (Dict[str, Any]): Context used to retrieve the information

Perform the backward pass for the multiplication operation. Args: ctx (Dict[str, Any]): Context used to retrieve the information

Perform the backward pass for the division operation. Args: ctx (Dict[str, Any]): Context used to retrieve the information for th

Perform the backward pass for the matrix multiplication operation. Args: ctx (Dict[str, Any]): Context used to retrieve the infor

Perform the backward pass for the flaten operation. Args: ctx (Dict[str, Any]): Context used to retrieve the information for the

Perform the backward pass for the conv2d operation. Args: ctx (Dict[str, Any]): Context used to retrieve the information for the

Perform the backward pass for the reshape operation. Args: ctx (Dict[str, Any]): Context used to retrieve the information for the

Perform the backward pass for the ReLU operation. Args: ctx (Dict[str, Any]): Context used to retrieve the information for the ba

Perform the backward pass for the reshape operation. Args: ctx (Dict[str, Any]): Context used to retrieve the information for the

Perform the backward step on the computational graph. Args: gradient (MPCTensor): The gradient (received) from the computational

Base for the FixedPrecision Encoder. Returns: int: base

Perform ABY3 bit decomposition for conversion of arithmetic share to binary share. Args: x (MPCTensor): Arithmetic shares of secr

Helper method that performs torch.cat on the replicated shares of the Tensors. Args: session_uuid_str (str): UUID to identify the session