Ransomware has rapidly become one of the most prevalent cybersecurity threats, often causing significant operational and financial disruptions by encrypting critical data and demanding ransom payments for decryption. Detecting ransomware early, particularly in network traffic, presents a unique challenge as many existing solutions rely on predefined malware signatures, which fail to detect new or evolving ransomware strains. A hybrid detection model combining Convolutional Neural Networks (CNN) for feature extraction and Isolation Forest for anomaly detection offers a novel and adaptive approach that does not depend on static signatures. The CNN component excels at identifying intricate traffic patterns, while the Isolation Forest algorithm efficiently isolates outliers, which correspond to ransomware activities. Extensive experimental evaluations demonstrate the model's superior performance in detecting unknown ransomware variants with high accuracy and low false-positive rates. The adaptability of the proposed solution makes it well-suited for real-time applications in dynamic network environments, offering a more effective defense mechanism against the fast-evolving landscape of ransomware threats.