This project develops a blockchain-based network for detecting fake news, leveraging NLP bases ML models and a consortium of auditors. The system operates on a consortium blockchain, enabling anonymous participation for readers to access news and submit news items as transactions.
- Public Network: Allows anonymous readers to read and broadcast news, maintaining privacy.
- Private Network: Comprises authenticated auditors who can propose blocks and vote on news validity.
- Auditor Joining: Requires authentication proof for registration.
- Bootstrap Nodes: Developer maintained nodes that are always active and used as entry points to the blockchain. They help discover peers in the network.
- Reader Joining: Simple registration with minimal information.
- News article transactions can be shared through a QR code, containing the IPFS url and metadata
- News Broadcasting: Users can upload and broadcast news, accompanied by a machine learning model score.
- Block Creation and Voting: Auditors propose and vote on blocks containing news transactions.
- Validation and Ledger Addition: Blocks receiving majority votes are added to the blockchain, making the news visible with associated authenticity metrics.
-
Clone the repository locally.
-
Change working directory into the repository folder
cd BlockchainProject2 -
Install dependencies by running the command (you can make a virtual environment if you want (recommended))
pip install -r requirements.txt -
Run the following command in your command line interface:
streamlit run screens/GUI.pyto start the blockchain project in your default browser.
-
Login/Sign up to continue
- When signing up, the email should be in the format username@domain.service
- Valid Certificate will be of the format ABCD****, for this POC.
-
If you receive an import error, kindly reload the page or run the commands in a new terminal. It is a streamlit bug rather than a code problem.
-
When you press "Go to Main", please wait for the server to initialise and connections. It will show a spinner & progress bar & take you to the main page automatically.
-
For viewing transactions in mempool or sent transactions, wait for the progress bar to load.
-
Similarly, please wait for the spinner showing "Please wait.." to finish before pressing buttons.
-
To upload a file, go to upload file page and upload a text file, attaching a transaction fee.
-
To view sent news, go to the sent news page. Similarly for viewing mempool and broadcasted blocks etc. The pages are present in the navbar.
-
The timer for the next block proposer is shown on every screen. Once this counter reaches 0, a new block proposer is chosen which is visible on the VIEW BLOCK STATUS page.
-
If you are chosen as the block proposer, you'll be shown the transactions in mempool on the VIEW BLOCK STATUS page to vote on them as "fake" news or not while showing the sender reputation, ML model score etc.
-
You can again view the current confirmations on the block in VIEW BLOCK STATUS.
-
You can view changes to your reputation in the REPUTATION LOG. The reasons for these changes may be broadcasting fake news, not broadcasting a block when chosen as a proposer, etc. (better described in report)
- Machine Learning Model: Utilizes a model trained on extensive datasets to assign a probability score indicating the likelihood of news being fake.
- Auditor Participation: Auditors, verified through a third-party service (e.g., Intel SGX), are responsible for block creation and news validation via voting.
- Reputation System: Differentiates between auditors and readers in terms of initial reputation scores, which are adjusted based on activity.
- Consensus Mechanism: Uses Proof of Stake (PoS) for secure and efficient consensus.
News items are displayed with their title, text, ML model score, percentage of fake votes by auditors, and sender reputation, providing a comprehensive view of their credibility.