/Spotify-Packet-Analysis

Primary LanguageJupyter Notebook

Spotify-Packet-Analysis

Analysis of Digital Audio Streaming Data

Final Report:

https://docs.google.com/document/d/14NxwLCIXSc70cwK8Ixc9oYui0Q0etwYuHvfFpOe2gqk/edit?usp=sharing

Objective:

Can we classify Spotify packets?

The high-level methodology of our research project encompasses the systematic collection of network data from the Spotify application and Web Browser. We will subsequently extract important features, apply classification models, and evaluate based on precision and recall as metrics.

Data:

.pcap and .csv

  • .pcap files will be processed using python (pcapkit)
  • raw data (direct features):
    • headers of protocols
    • source and destination IP
    • payload data of the packets
  • processed data (derivative features):
    • inter-arrival time between packets
    • average packet inter-arrival time per destination (Spotify server)

Tasks:

What mechanisms are we using

  • Collect network data using tcpdump and scapy
  • Feature engineering/extraction
  • Apply ML models to dataset

Evaluation:

How are we evaluating the ML models?

  • Compare Precision/Recall and RMSE
  • Confusion Matrix
Actual Class Classified as non-Spotify Classified as Spotify
non-Spotify data True Negative False Positive
Spotify data False Negative True Positive

Deliverables:

Generated outputs

  • bash/py script (that performs automated data collection)
  • Raw dataset (.pcap)
  • Clean dataset
  • ML model (Decision Tree, Random Forest, XGBoost)
  • Final report

File Structure

.
├── script                  # .py and .sh to collect packets
├── data                    # .pcap and .csv
├── notebook                # .ipynb for feature engineering, visualization and modeling
├── visualization           # .png or .pdf (saved visualizations)
├── .env                    # to save API keys
├── .gitignore
├── requirements.txt
├── environment.yml
└── README.md