A Deep Learning-Based Hybrid Model for Enhancing Security

The compliance of autonomous weapon systems (AWS) with International Humanitarian Law (IHL) is a pressing issue in contemporary warfare. This study proposes LawCompNet, a hybrid deep learning model designed to analyze the adherence of A WS operations to key IHL principles such as distinction, proportionality, and necessity. The model integrates a transformer-based Legal Context Understanding Module (using LegalBERT) and a Decision Analysis Module combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. Using a curated dataset of legal documents, treaties, and case law, the system achieved high performance with an accuracy of 92.4% and an F1-score of 91.9% on the test set. Class-wise analysis shows strong performance, particularly in distinction compliance (F1-score: 93.2<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>) and proportionality compliance (F1-score: 92.2<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>). Comparative evaluation with baseline models demonstrates the superior accuracy and interpretability of the hybrid approach. This research not only advances the use of AI in legal compliance analysis but also highlights the potential for enhancing decision-making frameworks for AWS. Future improvements include expanding datasets, addressing rare edge cases, and refining interpretability mechanisms to better align with evolving IHL standards.