Hi,

Can you release the source code of DeepSigns? Otherwise, we cannot run your examples (cannot import deepsigns package) and do further work.

Thanks

Hi, I have some question on the original paper:

"Authors in [11, 12] propose an N-bit (N ≥ 1) watermarking approach for embedding the IP information in the static content (i.e., weight matrices) of DL models."

"More recent studies in [13, 14] propose 1-bit watermarking methodologies for DL models."

What is the definition of the concept of N-bit? Why [13,14] are two 1-bit watermarking approaches?

import DeepSigns
Traceback (most recent call last):

File "", line 1, in
import DeepSigns

ImportError: bad magic number in 'DeepSigns': b'\x16\r\r\n'