Malware detection in blockchain using CNN

Abstract

The inherent decentralized nature and peer-to-peer system of the blockchain’s popularity has been on the rise in recent times and is being adopted in various innovative applications. This technology claims to be one of the most secure inventions due to the employment of hash functions, which makes the data stored immutable. However, security issues concerning blockchains have been highlighted in recent reports, which begs the question: is the blockchain technology as invulnerable as it once claimed to be? These reports talk about malware injections which lead to data corruption, data theft as well as third parties gaining networking power. This has become a significant worry for security in the dynamic online world. To counter such security concerns, we propose a model which combines a convolutional neural network with a blockchain in order to prevent malicious data transactions and thus malware injection within a blockchain network. This convolutional neural network detects any malware that might be present in the data before a new block is created to be a part of the blockchain. We have compared two different CNN models: the VGG-16 architecture and a customized model with fewer layers. When integrated with our blockchain model, the VGG-16 convolutional neural network architecture achieves an accuracy of 90.3% while the custom model achieves an accuracy of 88.90%.