ABC-Verify: AI-Blockchain Integrated Framework for Tweet Misinformation Detection

The propagation of misinformation has become prevalent in recent years and is one of the predominant factors for social media myths and conspiracy theories. This paper proposes and develops a solution to detect fake news and hence control misinformation broadcasting in social media. Existing solutions for detecting fake news involve either using Machine learning/AI or employing a crowdsourcing-based fact-checker to evaluate the reliability of the information. In our proposed solution - ABC-verify - we designed and developed an integrated framework combining both AI and a Proof-of-stake (PoS) smart contract algorithm for crowdsourcing to achieve better accuracy than AI-only or pure crowdsourcing. The advantage of the proposed solution is two-fold. Firstly, the AI model can continuously learn from the output of the smart contract algorithm. Secondly, the validated news that is added to the blockchain is immutable. The validators from the public Ethereum blockchain stake ERC721 tokens in exchange for a reward if the information reliability were accurate. The prediction from the AI classification model is based on a pre-trained BERT model on a dataset of 10,000 labelled Twitter datasets. The AI classification model proxies as one of the validators in the PoS algorithm. The final verdict from the smart contract is then fed back into the training dataset to improve the AI classification model and achieve better overall accuracy of 93%. Unlike traditional crowdsourcing platforms, news stored within the blockchain is immutable. Furthermore, the Ethereum blockchain is transparent, and every transaction is recorded within the blockchain, hence enabling authenticity and trust between peer-to-peer transactions.