The issue of copyright infringement is a significant threat to the creative freedom of composers, as unauthorized use of their work denies them recognition and compensation. Traditional methods like metadata-based systems and audio fingerprinting struggle with modified versions such as remixes, lo-fi adaptations, and nightcore variations. This research aims to address the problem of copyright infringement by developing a model that can detect unauthorized use of music tracks. The approach leverages wavelet-transform-based feature extraction and Long Short-Term Memory (LSTM) networks to analyze audio files. Key features such as statistical chroma features, MFCC coefficients, and zero-crossing rates are extracted from both original and modified audio files, including remixes, lo-fi, and nightcore versions. By training the LSTM model on these features, the study aims to distinguish between copyrighted and infringing music, demonstrating that advanced machine learning techniques can effectively identify unauthorized modifications. The outcomes of this research will contribute to the protection of intellectual property in the music industry by providing a robust method for detecting copyright violations in audio content.