fss::Dcf< in_bits, Group, Prg, In, pred, par_depth > Class Template Reference

2-party DCF scheme. More...

Classes
struct	Cw
	Correction word. More...

Public Member Functions
void	Gen (Cw cws[], const int4 s0s[2], In a, int4 b_buf)
	Key generation method.
int4	Eval (bool b, int4 s0, const Cw cws[], In x)
	Evaluation method.
void	EvalAll (bool b, int4 s0, const Cw cws[], int4 ys[])
	Full domain evaluation method.

Public Attributes
Prg	prg

Detailed Description

template<int in_bits, typename Group, typename Prg, typename In = uint, DcfPred pred = DcfPred::kLt, int par_depth = -1>
requires ((std::is_unsigned_v<In> || std::is_same_v<In, __uint128_t>) && in_bits <= sizeof(In) * 8 && Groupable<Group> && Prgable<Prg, 4>)
class fss::Dcf< in_bits, Group, Prg, In, pred, par_depth >

2-party DCF scheme.

Template Parameters

in_bits	Input domain bit size.
Group	Type for the output domain. See Groupable.
Prg	See Prgable.
In	Type for the input domain. From uint8_t to __uint128_t.
pred	See DcfPred.
par_depth	-1 is to use ceil(log(num of threads)), which should be good enough. Only EvalAll() uses it. See EvalAll() for details.

Member Function Documentation

Eval()

template<int in_bits, typename Group , typename Prg , typename In = uint, DcfPred pred = DcfPred::kLt, int par_depth = -1>

int4 fss::Dcf< in_bits, Group, Prg, In, pred, par_depth >::Eval ( bool  b, int4  s0, const Cw  cws[], In  x  )

inline

Evaluation method.

Parameters

b	Party index. False for 0 and true for 1. \(i\).
s0	Initial seed of the party.
cws	Returned by Gen().
x	Evaluated input. \(x\).

Returns

Output share. \(y_{i,x}\).

EvalAll()

template<int in_bits, typename Group , typename Prg , typename In = uint, DcfPred pred = DcfPred::kLt, int par_depth = -1>

void fss::Dcf< in_bits, Group, Prg, In, pred, par_depth >::EvalAll ( bool  b, int4  s0, const Cw  cws[], int4  ys[]  )

inline

Full domain evaluation method.

Evaluate the key on each input, i.e., 0b00...0 - 0b11...1. Store the outputs sequentially.

b, s0, cws are the same as the ones in Eval().

Parameters

ys	Pre-allocated output array. Its size must be at least 2 ** in_bits.

Support parallel using OpenMP. The task is divided to 2 ** par_depth parallel sub-tasks with the equal workloads. par_depth = -1: use ceil(log(num of threads)). par_depth = 0: no parallelism, i.e., sequential execution.

Gen()

template<int in_bits, typename Group , typename Prg , typename In = uint, DcfPred pred = DcfPred::kLt, int par_depth = -1>

void fss::Dcf< in_bits, Group, Prg, In, pred, par_depth >::Gen ( Cw  cws[], const int4  s0s[2], In  a, int4  b_buf  )

inline

Key generation method.

Parameters

cws	Pre-allocated array of Cw as returns. The array size must be in_bits + 1.
s0s	2 initial seeds. Users can randomly sample them.
a	\(a\).
b_buf	\(b\). Will be clamped and converted to the group element.

The key for party i consists of cws + s0s[i].

---

The documentation for this class was generated from the following file:

• include/fss/dcf.cuh