As smart home environments become increasingly reliant on a vast network of interconnected Internet of Things (IoT) devices, they face a growing number of cyber threats. These threats originate from both external sources, such as malicious actors on the internet, and internal sources, such as compromised or malfunctioning IoT devices within the home network. These threats can take various forms, including unauthorized access, Denial of Service (DoS) attacks, data exfiltration, and other malicious activities that can compromise the security and privacy of smart home inhabitants. To address this challenge, this paper presents an Intrusion Detection System (IDS) designed to enhance the security of IoT-based smart home networks. By leveraging a combination of Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) models, along with transfer learning and hyper-parameter optimization techniques, the proposed IDS efficiently detects both external network and intra-network smart home threats. Using the IoTID20 dataset, which simulates real-world attack scenarios such as Mirai botnet attacks, network scans, Denial of Service (DoS) attacks, and Man-in-the-Middle ARP spoofing, we trained and evaluated the IDS to recognize abnormal behavior within smart home networks. CNN layers are utilized to extract spatial features from network traffic, while LSTM layers capture temporal dependencies, providing robust detection capabilities against a range of attacks. The evaluation of the proposed IDS demonstrates high detection accuracy and F1-scores exceeding 99.45% on IoTID20 dataset, proving its effectiveness in safeguarding smart homes from evolving cyber threats.