Blockchain for Machine Learning Protetion Against Attacks in IoT

All sectors are adopting open innovations such as intelligent automation with the use of sophisticated technology such as artificial intelligence (AI) to accomplish systematic procedures for Internet of Things (IoT) systems. The practical deployment of AI-enabled smart systems is now limited by a lack of security levels and confidence in detecting IoT system threats. Because Blockchain can decrease AI vulnerabilities and AI can increase Blockchain's performance, these two technologies are complementary. In the literature, there is little research aimed at protecting intrusion detection systems (IDS) based on ML (Machine Learning) against evasion and poisoning attacks adopting, in most cases, statistical schemes based on traditional methods or ML and resulting in additional deployment and runtime costs. Although the power of machine learning tools is insufficient, these techniques are used to protect themselves against attacks. For this reason, we aim, in this article, to propose a new framework called intellig_block for the detection of cyber threats in ML models which will be used to develop IDS. In intellig_block, the execution of the classification technique will be decentralized, by hashing the model file and we integrate this hash as a smart contract. The experimental result shows encouraging results and low execution time and overhead in terms of gas consumed.