Living off the Land (LotL) attacks have gained attention in recent years because they are sneaky. These attacks exploit legitimate tools, scripts, and system permissions, making them hard to detect and track. As a result, defense costs increase. However, most research focuses on detecting and classifying malware rather than LotL attacks. This study aims to explore a novel approach that combines machine learning, deep learning, and natural language processing methods for detecting LotL attacks. We propose a deep learning detection framework called LOLWTC. It utilizes word embedding to represent qualitative features in command-line text as two-dimensional matrices. These matrices are then used for classification using deep networks. Experimental results demonstrate the robustness and effectiveness of LOLWTC. It achieves an impressive f1 score of 0.9945 on the test set during 10-fold cross-validation. This showcases its significant potential for detecting LotL attacks and addressing associated security concerns.