<small>_**Full-stack software: from network traffic capture to ML classification : identifying 35 applications and translating internet usage into real-time carbon emissions**_</small>
Streaming videos, sending emails, editing documents, every click **we make have a footprint**: how can we make it visible, measurable, and actionable?
I designed, tested, and deployed: the Internet Carbon Emission tool, a real-time software that helps people understand the environmental impact of their online activity.
Built on a **network traffic analyser** and a **machine-learning classifier**,the tool identifies up to **35 network applications** and translates that data into a clear, visual breakdown of each user’s internet footprint.
![[../assets/Attachments/internet/Pasted image 20251127152420.png]]
![[../assets/Attachments/internet/Pasted image 20251127152433.png]]
![[../assets/Attachments/internet/Pasted image 20251127152447.png]]
## the problem
The more people connect, our digital carbon footprint keeps growing quietly in the background.
- Internet traffic is projected to rise **26% every year** [1]
- It already accounts for **7% of global electricity use** and **3.5% of total GHG emissions** [1]
- Without change, internet traffic could reach **14% of global emissions by 2040**[3]
Two challenges stand in the way:
1. **It’s invisible.** Most users don’t realise that their data use produces carbon.
2. **It’s uncertain.** Current estimation tools vary wildly, sometimes by five orders of magnitude.
## how did i built it?
1. **Defining system boundaries** : This model adopts the widest possible system boundaries to capture the total GHG footprint. The results of this model can be segmented into sub-systems Data centers, Networks, User and Manufacturing and Fabrication.
![[../assets/Attachments/internet/internet.jpg]]
_The data for these subsystems is sourced from the International Energy Agency (IEA, 2022), as well as studies by Anders Andrae (Andrae, 2020)_ .
2. **Collecting the data:** The tool uses Wireshark (Network Traffic Analyzer) to capture Internet network traffic of users.
3. **Analysing and transforming the data:** A Random Forest ML model classifies 35 common applications and their data usage (GB) are converted into carbon emissions equivalent (CO₂e) based on location-specific energy grids.
![[../assets/Attachments/internet/carbon_formula_box_1.png]]
## References
- Aslan et al. (2018) - Electricity intensity of Internet data transmission.
- Andrae & Edler (2015) - Global electricity usage of communication technology.
- Greenpeace’s Click Clean Report (2016) - Breakdown of data traffic per application.
***Check my code*** [**Github**](https://github.com/cocoritzy/Internet-carbon-footprint)