ckksvector multiplied with identity matrix results in transparent ciphertext

[MartinNoc]

Question

Is it a bug or intended behavior when I get a "result ciphertext is transparent" error when multiplying a ckksvector with a numpy identity matrix?

Further Information

When multiplying a ckksvector with a numpy identity matrix, I get the error "result ciphertext is transparent" which seems very strange to me, as the result should be the same ciphertext as before, right? Moreover, when I compute the dot product of the ciphertext and any identity matrix column, it works.

System Information

• OS: Windows
• Python 3.9.7, tenseal 0.3.8
• Conda 4.11.0

Code

import tenseal as ts
import numpy as np
  
# Create TenSEAL context
bits_scale = 32
context = ts.context(
    ts.SCHEME_TYPE.CKKS,
    poly_modulus_degree=16384,
    coeff_mod_bit_sizes=[43] + [bits_scale]*11 + [43]
)
context.global_scale = 2**bits_scale
context.generate_galois_keys()

N = 100
x = np.random.rand(N)
eye = np.eye(N)
x_enc = ts.ckks_vector(context, x)

# this works
x_enc.dot(eye[:,0])

# this results in an error
x_enc @ eye

[MartinNoc]

Ok, I figured out, TenSEAL uses the diagonal method to compute vector-matrix products. This means for the identity matrix that all vectors except the first one are zero, resulting in the transparent ciphertext error.

See here the corresponding code piece.

How can we solve this? In my opinion, the transparent ciphertext should be thrown for the resulting product, not for intermediate results.