Crime Canary

What is Crime Canary?

Crime Canary is a framework that involves the Algorand block chain technology for the incentivization and automation of law enforcement reporting. Specifically, a user (civilian) can utilize the framework to report an incident, a crime, etc. to a local law enforcement via their submission of an image. Additionally, if the submission is proven to be valid, the user is then rewarded with governmental crypto-coins, which can be traded for approved services.

Hence, the overall goal is

• To create an automated, transparent, and incentivized platform for fast and accurate incident reporting
• To reduce law enforcement response times, which may help save lives, apprehend crime perpetrators, and generally improve the safety of citizens
• To increase the volunteer participation of citizens in law enforcement

This is a research project and not an official product of any governmental entities. Expect bugs and sharp edges.

Installation

Make sure you have python3 environment set up. Run the following to install the necessary packages for this project.

  pip3 install -r requirements.txt

Then, git clone the Algorand Sandbox (https://github.com/algorand/sandbox)

  cd sandbox
  ./sandbox up testnet

Usage

Launching the Web App

  python web_app.py

Creating Accounts

  python create_accounts.py

---

Alert Requirements

Users need to first opt-in to the police wallet, where default wallet-id is 176869006 and the default police account is listed in accounts.csv.

Then, users need to send a stake (or a transaction) to the police wallet first, where the attached message is the hash of the image they are submitting.

Finally, the users utilized the web-app to finalize their report.

For example, for demo purposes, we have a quick script to perform opt-in and staking.

python opt_and_stake.py CGQY5D62KVYRDWNHSY6MVLPMA7TBAHZ6I2PSOPZB7A27GIZEJU3C2XEXQ4 9d7QxuJMq08sNMXPVXZWYM/lSElr+KGkSoEk0TOTfNERoY6P2lVxEdmnljzKrewH5hAfPkafJz8h+DXzIyRNNg== ~/Desktop/test_image.jpg

This particular script opt_and_stake.py takes in the public key and private key of the user along with the path to an image.

Sequence Diagram

Model

Update (3.23.23)

1. Initial code for web app.
   • Users can submit image and their public key
2. Code for creating accounts on Algorand

Update (4.6.23)

1. Web app can load a model and predict an image

   • Still need extra functionality on ensuring the size of image is correct to the model

2. Web app can check if a user did put a stake on the blockchain

3. Siamese Network is added along with experimental results following the work on Convolution neural network joint with mixture of extreme learning machines for feature extraction and classification of accident images

Update (4.14.23)

1. Integrated the Oracle

2. Integrated discrete codes from different components to work together

3. Web app can validate an image (of any size), where the size of the image is ensured to follow a default shape required by the model.

4. Users are rewarded based on severity of the predicted accident.

Contributors

Chibuikem Ezemaduka (ezemac@rpi.edu)

Bao Pham (phamb@rpi.edu)

Maruf A. Mridul (mridum@rpi.edu)