Cyberattacks are very prevalent today. With the advent of IoT, many electronic devices such as smart fridges, watches, lights, cameras, and entertainment systems are all connected to the internet, which raises concerns about cybersecurity risks. We analyze IoT Dataset-23 in this study using conn.log Zeek files that contain 20 examples of malicious assaults and three examples of benign captures. We used the pandas library to combine these files into a single data frame, generating a single CSV file. Our methodology comprised the construction of two separate datasets: one for binary classification, which discerns between benign and harmful actions, and another for multi-class classification, which covers a range of malicious attack types. The dataset was initially analyzed and preprocessed to remove redundant data and ensure optimal model performance. To address class imbalance, the Edited Nearest Neighbors (ENN) and Synthetic Minority Over-sampling Technique (SMOTE) were used, which improved the representation of minority classes while retaining dataset integrity. Multiple deep learning (DL) and machine learning (ML) models were thoroughly tested on both multi-class and binary classification tasks. The best-performing model was the Long Short-Term Memory (LSTM) model, which achieved outstanding results with an accuracy rate of 99.7%, a recall rate of 97%, an F1 score of 98%, and a precision rate of 100%.