In modern diplomacy, secure and truthful communication is paramount, with encrypted messaging increasingly employed to protect sensitive international exchanges. This paper presents a two-stage deep learning approach to tackle the dual challenges of analyzing encrypted diplomatic messages: identifying the encryption algorithms applied and detecting deceptive intent in negotiations. Utilizing the Diplomacy dataset, which consists of annotated pairwise diplomatic conversations labeled for truthfulness and deception, we first conduct cryptanalysis to classify the encryption algorithms employed, specifically focusing on AES-128, Camellia-128, Twofish-128, and the Simon & Speck block cipher algorithms. Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) are leveraged to capture distinct features and sequential patterns from the encrypted messages generated by these algorithms. In the second stage, we employ Bidirectional Long Short-Term Memory (BiLSTM) networks and hybrid CNN-LSTM models to assess the sender’s negotiation intent solely based on the encrypted text. Our findings demonstrate the effectiveness of combining cryptanalysis with deception detection, advancing automated analysis in encrypted diplomatic communications and providing insights into secure communication in high-stakes, trust-sensitive environments.