The popularity of Windows attracts the attention of hackers/cyber-attackers, making Windows devices the primary target of malware attacks in recent years. Several sophisticated malware variants and anti-detection methods have been significantly enhanced and as a result, traditional malware detection techniques have become less effective. This work presents MalBehavD-V1, a new behavioural dataset of Windows Application Programming Interface (API) calls extracted from benign and malware executable files using the dynamic analysis approach. In addition, we present MalDetConV, a new automated behaviour-based framework for detecting both existing and zero-day malware attacks. MalDetConv uses a text processing-based encoder to transform features of API calls into a suitable format supported by deep learning models. It then uses a hybrid of convolutional neural network (CNN) and bidirectional gated recurrent unit (CNN-BiGRU) automatic feature extractor to select high-level features of the API Calls which are then fed to a fully connected neural network module for malware classification. MalDetConv also uses an explainable component that reveals features that contributed to the final classification outcome, helping the decision-making process for security analysts. The performance of the proposed framework is evaluated using our MalBehavD-V1 dataset and other benchmark datasets. The detection results demonstrate the effectiveness of MalDetConv over the state-of-the-art techniques with detection accuracy of 96.10%, 95.73%, 98.18%, and 99.93% achieved while detecting unseen malware from MalBehavD-V1, Allan and John, Brazilian, and Ki-D datasets, respectively. The experimental results show that MalDetConv is highly accurate in detecting both known and zero-day malware attacks on Windows devices.