Advanced social media crime prevention via deep learning and cryptographic data encryption

Abstract The increasing prevalence of crime facilitated through social media platforms has become a critical concern for law enforcement and security agencies. With vast amounts of personal, sensitive, and actionable data being shared online, the risk of criminal exploitation has grown significantly, necessitating innovative approaches to crime prevention. This research addresses the problem of detecting and preventing social media-based crimes by integrating cryptographic data encryption with a focus on privacy and security. This study proposes a hybrid ECC-RSA (Elliptic Curve Cryptography—Rivest, Shamir, Adleman) model with key selection optimized using the Zebra Optimization Algorithm (ZOA) to enhance both the security and efficiency of the encryption process. Additionally, an Artificial Neural Network (ANN) is employed as a specific application to classify and detect suspicious online activities related to criminal behavior. A key challenge in this domain is achieving a balance between privacy preservation and effective crime detection. To evaluate the system, an extensive dataset of social media interactions containing both legitimate and suspicious activities was used. The results demonstrate that the integrated ECC-RSA model with ZOA optimization provides robust encryption while maintaining high detection accuracy, achieving 92% accuracy, 89% precision, and 91% recall. The ANN-based detection system successfully identifies potential criminal activities, while the cryptographic model ensures no sensitive data is exposed during analysis, maintaining user privacy. The findings suggest that the proposed hybrid model offers a promising solution for proactive crime prevention on social media, effectively balancing privacy, security, and detection performance.