Sequre features

Sequre operates on top of tensors of arbitrary dimension. Specifically, the SharedTensor class stores the secret additive share and auxiliary data, such as Beaver partitions, as n-dimensional arrays at each computing party. Furthermore, the compile-time optimizations apply only to SharedTensor expressions. The table below presents all secure operations supported for SharedTensor operands.

Secure operation	Example usage (x and y are SharedTensors)
Addition / Subtraction	x + y; x - y
Element-wise multiplication	x * y
Exponentiation by a constant	x ** c (c is a constant)
Element-wise division	x / y
Comparisons	x == y; x != y; x > y; x < y; x >= y; x <= y;
Square root	sqrt(x)
Dot product / Matrix multiplication	dot(x, y); matmul(x, y)
Element-wise absolute value	abs(x)
Max/min element (vector only)	max(x); min(x)
Argmax/argmin element (vector only)	argmax(x); argmin(x)
Householder transformation (matrix only)	householder(x)
QR factorization (matrix only)	qr_fact_square(x)
Tridiagonalization (matrix only)	tridiag(x)
Eigen decomposition (matrix only)	eigen_decomp(x)
Orthonormal basis (matrix only)	orthonormal_basis(x)
Linear SVM training, predicting, scoring	lsvm_train(x, y, …), lsvm_predict(x, …), lsvm_score(x, y, …)
Element-wise bit decomposition	bit_decomposition(x, …)
Element-wise bit-wise addition	bit_add(x, y)