ENHANCED AI SECURITY WITH DWT WATERMARKING AND HYBRID ANOMALY DETECTION FRAMEWORK (HADF)

Abstract

This paper presents a novel approach to enhancing security in artificial intelligence systems through the fusion of Discrete Wavelet Transform (DWT)--based watermarking with a Hybrid Anomaly Detection Framework (HADF). Traditional watermarking techniques often struggle to withstand various attacks in digital environments, especially in the context of AI systems where the stakes are high. In response, the proposed framework combines the robustness of DWT-based watermarking with the adaptive capabilities of anomaly detection to create a more resilient security mechanism. The DWT-based watermark embeds imperceptible information into the host data, serving as a unique identifier for authentication and ownership verification. Meanwhile, the Hybrid Anomaly Detection Framework leverages machine learning algorithms to continuously monitor system behavior, detecting and responding to anomalous activities in real time. By integrating these components, the proposed framework not only enhances the security of AI systems but also ensures their integrity and reliability in the face of evolving threats. Experimental results demonstrate the effectiveness of the approach in detecting and mitigating attacks while maintaining system performance and usability. Overall, the fusion of DWT-based watermarking with the Hybrid Anomaly Detection Framework offers a promising solution for bolstering security in AI systems, thereby fostering trust and confidence in their deployment across various domains.