Go to AI Workflows

1 Functions 2 Reference Architecture 3 I/O Data
4 Functions of AI Modules 5 I/O Data of AI Modules 6 AIW, AIMs, and JSON Metadata
7 Reference Software 8 Conformance Texting 9 Performance Assessment

1        Functions

The With Inference Robustness (NNW-WIR) receives watermarked parameters, the testing dataset, the payload and provides both the retrieved payload and the number of incorrect bits (Count error).

2        Reference Model

Figure 1 specifies the With Inference Robustness (NNW-WIR) Reference Model including the input/output data, the AIMs, and the data exchanged between and among the AIMs.

Figure 1 – Reference Model of With Inference Robustness (NNW-WIR))

The operation of With Inference Robustness (NNW-WIR) develops in the following way:

  1. A user provides
    1. The Original payload
    2. The Testing dataset
    3. The Watermarked parameters
  2. The machine provides
    1. The Count Error
    2. The Retrieved Payload

3         I/O Data

The input and output data of the With Inference Robustness (NNW-WIR) Use Case are:

Table 1 – I/O Data of With Inference Robustness (NNW-WIR)

Input Descriptions
Original Payload The information originally inserted.
Testing dataset A set of input to do the inference.
Watermarked parameters The parameters of a watermarked AIM.
Output Descriptions
Count error The number of incorrect bits in the retrieved payload.
Retrieved payload The output of the decoding procedure of the watermarking method.

4        Functions of AI Modules

Table 2 provides the functions of the With Inference Robustness (NNW-WIR) Use Case.

Table 2 – Functions of AI Modules of With Inference Robustness (NNW-WIR)

AIM Function
Modification module Modifies the parameters of the watermarked AIM.
Modified AIM Produces inference using the testing dataset and the modified parameters.
WIR Watermark Decoder Retrieves the payload using the watermarking method.
Comparator Compares the retrieved payload to the original payload.

5         I/O Data of AI Modules

The AI Modules of With Inference Robustness (NNW-WIR) are given in Table 3.

Table 3 – AI Modules of With Inference Robustness (NNW-WIR)

AIM Receives Produces
Modification module Watermarked parameters Modified parameters
Modified AIM 1.     Testing dataset

2.     Modified parameters

 Modified inference
WIR Watermark Decoder Modified inference Retrieved payload
Comparator 1.     Original payload

2.     Retrieved payload

 Unwatermarked inference

6        AIW, AIMs, and JSON Metadata

Table 4 provides the links to the AIW and AIM specifications and to the JSON syntaxes. AIMs/1 indicates that the column contains Composite AIMs and AIMs indicates that the column contains their Basic AIMs.

Table 4 – AIW, AIMs, and JSON Metadata

AIW AIM Name JSON
NNW-WIR With Inference Robustness X
NNW-MFM Modification Module X
NNW-MDM Modified Module X
NNW-WWD WIR Watermark Decoder X

7        Reference Software

7.1       Disclaimers

  1. This NNW-WIR Reference Software Implementation is released with the BSD-3-Clause licence.
  2. The purpose of this Reference Software is to demonstrate a working Implementation of NNW-WIR, not to provide a ready-to-use product.
  3. MPAI disclaims the suitability of the Software for any other purposes and does not guarantee that it is secure.
  4. Use of this Reference Software may require acceptance of licences from the respective repositories. Users shall verify that they have the right to use any third-party software required by this Reference Software.

7.2        Guide to the NNW-WIR code

Use of this AI Workflow is for developers who are familiar with Python and PyTorch libraries,

The robustness.py code allow a User to evaluate the robustness of a watermarking method on the image classification task:

  1. The watermarking method is implemented as a Python Class
  2. The attack is performed using mainAttack.py

The NNW-WIR Reference Software is found at the gitlab site. It contains:

  1. The python code implementing the AIW.
  2. The required libraries are: pytorch, tqdm

8        Conformance Testing

9        Performance Assessment

Go to AI Workflows