DL-based Framework for Detecting Malicious Proof-of-Stake Blocks in Gaming Transactions

Abstract

Blockchain gaming platforms based on Non-Fungible Tokens (NFTs) have vastly grown and provide players with ownership to trade and monetize in-game assets. However, the integrity of NFT transactions in Proof-of-Stake (PoS) systems are prone to security threats due to the possible insertion of malicious blocks. Detection of these blocks is required to keep trust in the ecosystem. The proposed machine learning-based solution included detection of malicious blocks in PoS with focus on the performance of an LSTM Model. Following extensive testing with 3 models of LSTM, 1D CNN and GRU, the model achieved an accuracy of 86.9% making it most effective in regards to early identification of fraudulent transactions. This method augments the security of NFT gaming transactions and thus offers support against several malicious acts occurring in blockchain based ecosystems.