Driver identification is crucial for automotive security, law enforcement, and ride-sharing industries, as well as for advanced driver assistance systems, fleet management, and usage-based insurance. Machine learning and deep learning techniques show promise for accurate identification, but a comprehensive analysis of existing methods is lacking. This paper addresses the gap by reviewing and analyzing existing techniques, including preprocessing, feature extraction, classification algorithms, and deep learning architectures. Performance, advantages, and limitations are critically evaluated. A future framework for driver identification with Large Language Model (LLM) is proposed, exploring its potential in this domain. Traditional methods like Support Vector Machines and Random Forests offer reliable performance, while deep learning requires larger datasets and computational resources but achieves higher accuracy. Synergies between established techniques and emerging technologies like LLM are identified for future research. Key directions include hybrid approaches and transfer learning for efficient adaptation to new datasets. This review serves as a valuable resource for researchers and practitioners, highlighting strengths, weaknesses, and promising research directions for highly accurate driver identification systems.