Procrastination Tracker and Goal Reminder

Maneesh Kumar Singh (mksingh4@illinois.edu)

This is source code repository for my final project for UIUC graduate course CS498 Internet of Things. As part of this project, I built a time management system which consists of two parts: A social media usage tracker with deterrent and a live display of current tasks or goals in hand using Raspberry Pi. We are using AWS to host our IoT infrastructure.

Project Resources

• Project Report
• Video URL. Video URL is only accessible using UIUC net id.

Code structure

tcpdump

tcpdump.sh

A bash script to run tcpdump with correct parameters to capture network traffic on wlan0 interface and write them to /var/tmp/pcap location.

mqtt_client.py

An AWS IoT client written using AWS IoT Python SDK. It has methods implemented for connect, disconnect, publish and subscribe.

mqtt_publish.py

This python module parses the pcap files created by tcpdump utility. It build a reverse DNS lookup dictionary using DNS records and uses them for reverse DNS lookup on subsequent network records. It summarizes and sends message to IoT topic in batches of 500 messages in a JSON array. Reverse DNS dictionary function is written by Je-Clark at his GitHub Repository.

lambda

This folder contains the source code for three lambda functions used in this project.

flatten_records

This lambda function is used by AWS IoT Analytics pipeline to flatten the JSON array send by mqtt_publish.py. It is based on code provided in AWS support forum.

leetcode_notification

This lambda function pulls the information from LeetCode and decides the next problem to publish to IoT topic. It requires access to LeetCode session cookie which can be stored in parameter store and retrieved by lambda.

notify_mentor

This lambda function maintains a global dictionary to track network data usage. It maintains a blacklisted website list and threshold. Once a blacklisted domain crosses the threshold it publishes a message to AWS SNS which in turns sends SMS to people subscribed for SMS delivery mode.

electron-app

This folder contains source code for Electron application used to display LeetCode status and allows user to request refresh by click.

How to run

We have developed and tested the code on Ubuntu and Raspberry Pi. Therefore we would provide instructions for Ubuntu and Pi only.

Setup

Please setup and activate virualenv for python >= 3.7.3. Clone the repository and then from src folder use pip to install all the necessary packages.

cd src
pip install -r requirements.txt

AWS IAM setup

Create new IAM roles or modify existing one to provide

• AWS Lambda role needs access to iot:Publish, sns:Publish and ssm:GetParameter
• AWS EC2 role for access to IoT Analytics dataset through IoT notebook. I used create new role option during creation of notebook instance.

AWS SSM setup (Optional)

Login to LeetCode using Google Chrome. Copy crsftoken cookie value from Developer tools and store it in AWS SSM parameter store as a secret value. Note down parameter path.

AWS SNS setup

Create an AWS SNS topic and note down the ARN. Also subscribe to topic using SMS delivery mode.

AWS Lambda function

If you have setup AWS SSM to store LeetCode cookie value, change the SSM path in leetcode_notification lambda_function.py module. leetcode_notification lambda requires extra libraries not available in AWS Lambda environment.

• Install dependencies in the leetcode_notification directory.
• zip the package

# From src/lambda/leetcode_notification
pip install -r requirements.txt -t .
zip -r leetcodenotification.zip

Modify notify_mentor lambda code with new SNS ARN generated in previous step. Other two lambda function do not require any external dependencies so you can zip them without install step.

# From src/lambda/
cd flatten_records
zip -r flatten_records.zip .
cd ../notify_mentor
zip -r notify_mentor.zip .

Create corresponding empty lambda functions with Python 3.7 runtime and upload the zip files. In AWS Lambda role, provide access to iot:Publish, sns:Publish and ssm:GetParameter. Change the Memory to 256 MB and runtime to 1 minute for all functions using configuration tab in AWS Lambda console.

AWS IoT Analytics setup

1. Please create an IoT analytics channel, data source, pipeline and query dataset. While creating pipeline, add AWS Lambda activity and configure FlattenRecords lambda created above.

2. Create an IoT Analytics notebook. It would requires an notebook instance to spin-up. You can create new role to provide access to notebook instance or you can use existing role if setup with right permissions.

AWS IoT setup

1. Please create a device in AWS IoT Core and download the certificates and keys to folder src/tcpdump/certificates. Find AWS IoT endpoint in Settings Page of AWS IoT Core console. Change the corresponding values in mqtt_publish.py

       client = MQTTClient(
           'xxxxxxxxxxxx-ats.iot.us-east-2.amazonaws.com',
           'certificates/d6ccf7c6bd-certificate.pem.crt',
           'certificates/d6ccf7c6bd-private.pem.key',
           'certificates/AmazonRootCA1.pem')
       client.connect()

2. Create an IoT rule to trigger AWS IoT analytics channel and AWS Lambda NotifyMentor for topic cs498/time/usage.

3. Create an IoT rule to trigger AWS Lambda LeetCodeNotification for topic cs498/time/new.

Network capture

Network capture part can be tested on Ubuntu or Pi without setting up Pi as network router. In this mode, it will capture the instance data flowing through wlan0 interface.

In order to deploy this system, you need to setup Raspberry Pi as an access point and network router. Please follow project report for complete instructions.

Run the tcpdump.sh script provided at src/tcpdump/tcpdump.sh. Please change the interface name or output location if different than as specified in the script.

cd src/tcpdump/
./tcpdump.sh

This will star capturing network traffic in /var/tmp/pcap folder.

Python parsing

Run mqtt_publish.py python module from src/tcpdump folder.

# from src/tcpdump
python mqtt_publish.py

This module would parse the network packets, summarize them and publish them to topic cs498/iot/usage.

If everything is correctly setup, you should get SMS on subscribed cellphone on high usage. You can play with threshold in NotifyMentor lambda to test value that suits your use.

Electron-app

Go to src/electron-app folder. Install npm dependencies and start the application.

cd src/electron-app
npm install
npm start

Once application starts you can click on Get New button to request information or wait for five minutes to get the message from IoT topic.

Acknowledgement

• Goal tracker is my personal favorite and was on my agenda to build something like for a long. I am thankful for Professor Matt to provide me an opportunity to implement this idea while learning about IoT systems and technologies.
• I would also like to acknowledge the support from TA and fellow classmates in completing this project and other course labs.
• Je-Clark for his work on ScapyDNSDissection at GitHub Repository.