Adaptive Deep Learning Architectures for Dynamic Cyber Forensic Analysis

The increasing complexity of modern cyber threats has highlighted the limitations of traditional forensic investigation methods that rely on static analysis and post-incident evidence interpretation. This study aims to develop an adaptive deep learning–based framework capable of supporting real-time cyber forensic analysis through continuous monitoring of heterogeneous telemetry data in distributed enterprise environments. The framework incorporates reinforcement learning–driven policy optimization to dynamically adjust classification thresholds during runtime. Experimental evaluation indicates that the proposed framework achieves a detection accuracy of 96.5%, with precision and recall values of 95.8% and 96.1%, respectively. In comparison with baseline convolutional neural network–based models, the system demonstrates a 5.2% improvement in classification accuracy and a 3.9% reduction in false-positive rate. Additionally, detection latency was reduced from 142 ms to 102 ms, enabling near real-time forensic response under dynamic threat conditions.

Ҳали таржима қилинмаган