This code was designed to run on the free5GC 5G core and its detailed description is located below.
As the execution environment has many components, the tested working versions of the main ones are listed down below:
- python: 3.8.10 (works with 3.7.x too)
- pip: 23.3.1
- tensorflow: 2.13.1
- keras: 2.13.1
- flask: 3.0.0
- go: 1.18.10
- free5GC: v3.3.0
- Ubuntu Server 20.04.3
NOTE: List updated on December, 2023
Please, check this other file here
The authors of [Kim et al. 2022] implemented the NWDAF module and its submodules (MTLF and AnLF) integrated to free5GC, however, they used an image dataset as their ML functionality.
First, a reprodction of [Kim et al. 2022]'s work was made (this README details the environment used in this process). After that, another ML functionality closely related to Computer Networks field was implemented. Instead of using an image dataset, a packet capture dataset containing 6 captures of 1000 packets each was created. This dataset was used to test the new ML functionality and the instructions to reproduce the environment used for this second phase are located on this other file here.
Currently, the integration between [Kim et al. 2022]'s NWDAF and our ML functionality isn't finished yet, so Keras and TesorFlow are not being used on our experiment.
Detailed instructions won't be added here as it isn't in the scope of this document, however as a general advice you should install the free5GC project and then follow these instructions configuring the IP address on AMF, SMF and UPF configuration files.
TIP: You should change the loopback IPs (127.0.0.x) to the one used by the LAN interface
NOTE: This other repository may help you get started with the free5GC setup
NOTE: In this section the commands are supported on a BASH console
- Install Python3 and pip
sudo apt install python3 python3-pip
pip install --upgrade pip
- Install Python required packages
pip install -r requirements.txt
- Install Flask
pip install flask
- Install TensorFlow
pip install tensorflow
NOTE: The interoperability between the NWDAF and the ML functionality has not been completed yet, so if you are using a constrained computational environment, you can skip installing tensorflow and it's dependencies
- Add Go Lang support
NOTE: If Go version specified on the first section is already installed, skip this step
TIP: Check Go version using the command go version on console
If another version of Go was previously installed, remove the existing version and install Go 1.18.10 using:
sudo rm -rf /usr/local/go
wget https://dl.google.com/go/go1.18.10.linux-amd64.tar.gz
sudo tar -C /usr/local -zxvf go1.18.10.linux-amd64.tar.gz
If not, install Go using the commands below:
wget https://dl.google.com/go/go1.18.10.linux-amd64.tar.gz
sudo tar -C /usr/local -zxvf go1.18.10.linux-amd64.tar.gz
mkdir -p ~/go/{bin,pkg,src}
# The following assumes that your shell is BASH:
echo 'export GOPATH=$HOME/go' >> ~/.bashrc
echo 'export GOROOT=/usr/local/go' >> ~/.bashrc
echo 'export PATH=$PATH:$GOPATH/bin:$GOROOT/bin' >> ~/.bashrc
echo 'export GO111MODULE=auto' >> ~/.bashrc
source ~/.bashrc
cd ~/free5gc/
git clone -b mnc_Public-5G https://github.com/oliveiraleo/mnc_NWDAF.git
cp mnc_NWDAF/nwdafcfg.yaml config/.
First, change the working directory to the NWDAF's one
cd ~/free5gc/mnc_NWDAF/nwdaf
Then load the packages below
go mod download github.com/antonfisher/nested-logrus-formatter
go get nwdaf.com/service
go mod download github.com/free5gc/version
cd ~/free5gc/mnc_NWDAF/nwdaf
go build -o nwdaf nwdaf.go
cd ~/free5gc/
cp -r mnc_NWDAF/nwdaf .
Execute the components in the following order:
- 5G Core (free5GC)*
- go module (nwdaf executable compiled on the previous section)
- python module
- temp_requester
* NRF must be running
cd ~/free5gc/nwdaf/
./nwdaf
# OR
go run nwdaf.go
Now the NWDAF should be running
cd ~/free5gc/nwdaf/pythonmodule/
python main.py
cd ~/free5gc/nwdaf/Temp_Requester/
go run temp_requester.go
After that, you should select your number:
- If "1" is selected, MTLF (model training function) is executed.
- Otherwise, "2" is selected, AnLF (analytics function) is executed.
- Then, you can try to select a number which means the dataset number.
Currently, the EMNIST dataset which is in the python module is being used. In Temp_Requester, the image is not transmitted (the data number is transmitted using the json).
On another machine (different from the free5GC one), run:
cd ~
git clone https://github.com/aligungr/UERANSIM
cd UERANSIM
sudo apt install make g++ libsctp-dev lksctp-tools iproute2
sudo snap install cmake --classic
make
cd ~/UERANSIM
nano config/free5gc-gnb.yaml
Find the section below:
ngapIp: 127.0.0.1 # gNB's local IP address for N2 Interface (Usually same with local IP)
gtpIp: 127.0.0.1 # gNB's local IP address for N3 Interface (Usually same with local IP)
And change 127.0.0.1 to the LAN IP from UERANSIM's machine
Now, on this line:
# List of AMF address information
amfConfigs:
- address: 127.0.0.1
port: 38412
Change 127.0.0.1 to the LAN IP from free5GC's machine
Use the webconsole for that, the detailed instructions are located here
cd UERANSIM
# gnb
build/nr-gnb -c config/free5gc-gnb.yaml
# ue
sudo build/nr-ue -c config/free5gc-ue.yaml
To install the virtual test environment, please, refer to the file VMs-setup.md
For more information regarding this second phase, please refer to this section and to the file linked above.
After completing the installation of the environment and capturing packages into files you are ready to begin with the Machine Learning tests
To reproduce our results, just follow the instructions below
If not done yet, clone the repository using
git clone -b mnc_Public-5G https://github.com/oliveiraleo/mnc_NWDAF.git
cd mnc_Public-5G
Create and load a Python virtual environment
python -m venv pyvenv
source pyvenv/bin/activate
pip install -r requirements.txt
NOTE: The interoperability between the NWDAF and the ML functionality has not been completed yet, so if you are using a constrained computational environment, you can skip installing tensorflow and it's dependencies
Then create the results folder
cd ML_test_code/
mkdir results
cd ML_test_code/
python nwdaf_ml.py
NOTE: The current state of the implementation uses relative paths for interacting with the files on the HDD, this is why you need to enter the ML_test_code folder
Now it's only a matter of using the menus to guide the execution. Keep in mind that the two possible flows of execution should be
- I- Train and save model > Inference; or
- II- Load model > Inference
NOTE: The model file is overwriten each time training is executed. The same behavior applies to the files under the results folder for each run of the script
Utilizing the implemented environment is straightforward. To incorporate your custom data, place it within the dataset folder, with training and inference data allocated to their respective folders bearing the same names
Following the data transfer, if necessary, modify the labels list on the 53rd line of the nwdaf_ml.py file. Then run the commands from the section above
If you used our dataset or found this repository useful, please, cite it as:
L. A. de Oliveira, E. Franco Silva and M. A. Ribeiro Dantas, "A NWDAF Study Employing Machine Learning Models on a Simulated 5G Network Dataset," 2024 IEEE Symposium on Computers and Communications (ISCC), Paris, France, 2024, pp. 1-6, doi: 10.1109/ISCC61673.2024.10733717.,
Or use the BibTex below:
@INPROCEEDINGS{10733717,
author={{de Oliveira}, Leonardo Azalim and Franco Silva, Edelberto and Ribeiro Dantas, Mário Antônio},
booktitle={2024 IEEE Symposium on Computers and Communications (ISCC)},
title={A NWDAF Study Employing Machine Learning Models on a Simulated 5G Network Dataset},
year={2024},
pages={1-6},
keywords={Adaptation models;Data analysis;Accuracy;Protocols;5G mobile communication;Computational modeling;Machine learning;Reproducibility of results;3GPP;Testing;5G;NWDAF;Network Traffic Characterization;Classification},
doi={10.1109/ISCC61673.2024.10733717}
}
Or in Portuese:
Oliveira, L., Silva, R., Lima, P., Pereira, A., Valadares, J., Silva, E., & Dantas, M. (2024). Análise da Funcionalidade da NWDAF no Core 5G Sobre um Conjunto de Dados. In Anais do XLII Simpósio Brasileiro de Redes de Computadores e Sistemas Distribuídos, (pp. 798-811). Porto Alegre: SBC. doi:10.5753/sbrc.2024.1474
@inproceedings{sbrc,
author = {Leonardo Oliveira and Rodrigo Silva and Pedro Lima and Antônio Pereira and Júlia Valadares and Edelberto Silva and Mário Dantas},
title = {Análise da Funcionalidade da NWDAF no Core 5G Sobre um Conjunto de Dados},
booktitle = {Anais do XLII Simpósio Brasileiro de Redes de Computadores e Sistemas Distribuídos},
location = {Niterói/RJ},
year = {2024},
issn = {2177-9384},
pages = {798--811},
publisher = {SBC},
address = {Porto Alegre, RS, Brasil},
doi = {10.5753/sbrc.2024.1474},
url = {https://sol.sbc.org.br/index.php/sbrc/article/view/29836}
}
The original code from upstream did not explicitly specify any license terms. However, the work contained in this repository is licensed under the GPLv3, as indicated in the LICENSE file, which is reflected in the notice provided below:
This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License version 3 as published by the Free Software Foundation.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.