While field of addiction is evolving, lack of validated diagnostic tools and biomarkers for behavioral addictions has created a need for innovative approaches. This research explores the use of Electroencephalography (EEG) and deep learning techniques to detect pornography addiction. Four distinct deep learning models were implemented, each with a unique architecture, including combinations of Long Short-Term Memory (LSTM), Gated Recurrent Units (GRU), and Attention mechanism. These models were designed to capture temporal dependencies in EEG data and analyze sequential patterns associated with pornography addiction. To assess their performance, the models were evaluated using Leave-One-Out (LOO) and train-test split (TTS) techniques. LOO preserved the temporal integrity of EEG sequences, and simulate real-life scenario, while TTS provided standard evaluation results. The proposed models demonstrated promising results, with an average accuracy of 94 % -97 % using TTS, and 45% to 49% for LOO, outperforming baseline models. This research contributes for laying the foundation for timely diagnosis and intervention.