Improved Attack Detection and Analysis in IoT Environment Data Derived from Intrusion Detection Systems

The interconnection and communication capabilities of many smart gadgets are made possible by the Internet of Things. Despite the practical advantages of the Internet of Things (IoT), it is susceptible to new forms of assaults due to the introduction of new technology. Intrusion detection systems (IDSs) need a lot of resources to determine the kind of assaults due to the complicated networks and vast amounts of data generated by the Internet of Things (IoT). In contrast, most intrusion detection methods are impractical for IoT networks due to the fact that these networks need greater computational resources for attack detection, which are in short supply on IoT devices. As a result, you need an intrusion detection system (IDS) that isn't too heavy yet can still detect emerging threats. In contrast, most intrusion detection methods are impractical for IoT networks due to the fact that these networks need greater computational capacity for attack detection, which are in short supply on IoT devices. As a result, you need an intrusion detection system (IDS) that isn't too heavy yet can still detect emerging threats. This study recommends a combination of Long Short-Term Memory (LSTM) and Random Forest to optimize detection performance while simultaneously reducing network complexity via the efficient elimination of unnecessary features. Applying hold-out, Stratified k-fold cross-validation, as well as percentage split test mode on the CICIDS-2017 dataset MachineLearning CSV version, the suggested IDS is verified with testing and training data. We opted for this dataset since it contains actual data from IoT networks. The suggested hybrid IDS shows promising experimental results for performance enhancement. Recall values average out to 1.000, resulting in an accuracy rating of 99.9%.