Enhanced Criminal Suspect Identification Using a Novel Smart System with Hybrid Encryption and ANN Classification

Abstract

The current advancement in the growth of crime rates and the complexity of crimes that are being committed in society call for the adoption of better methods of identifying criminal suspects. While traditional methods are somewhat useful, they are not sufficient in dealing with the intricacies and the sheer scale of today’s data landscape. Machine intelligence has become popular in many fields including criminal justice because of its capability to learn from big data. This study presents a novel smart system for classifying criminal suspects using four key steps: query-based authentication (QBA), data categorization, data encryption and decryption, and artificial neural network (ANN)-based classification. QBA ensures only authorized access to sensitive data by verifying user-specific information. Data is categorized into sensitive (personal and social criminal data) and non-sensitive (classification results) categories, with sensitive data encrypted using a Two-Level hybrid ECC (Elliptic Curve Cryptography) and ECC-RSA (Rivest, Shamir, Adleman) model, optimized via the HSMEO algorithm for high security and efficiency. The ECC-RSA model outperforms traditional encryption methods (AES, DES, RSA, ECC) in security (98.53%), trust score (4.83), throughput (77.57), and encryption/decryption times (3.459/2.994 seconds). Additionally, the HSMEO model significantly reduces key generation time to 1.97545 seconds, surpassing other optimization strategies like SMO, EO, PSO, MFO, and FFO. Graphical representations of key metrics validate the ECC-RSA model’s superior performance in security, efficiency, and reliability, making it an effective method for protecting sensitive data and ensuring efficient criminal suspect classification through expert system.