AI-Assisted ISP and Chip-Off Forensic Framework for Damaged Android Devices

Physical damage to smartphones creates a persistent bottleneck in mobile forensic practice: once a device can no longer be accessed through its operating system, conventional logical acquisition fails, and investigators face a choice between accepting data loss and escalating to hardware-level intervention. This paper describes an integrated forensic workflow that addresses this gap by combining In-System Programming (ISP) and Chip-Off memory extraction with an AI-assisted artifact localization and prioritization layer. The workflow was evaluated on 18 physically damaged Android smartphones for which all standard acquisition paths were unavailable. Hardware extraction produced verified binary memory images from all 18 devices. A 1D-CNN localization classifier subsequently screened those images, achieving F1-score = 0.88 and ROC-AUC = 0.94 on the synthetic test partition. Prioritization of candidate windows reduced manual review volume by 78%, cut total expert review time by 63%, and shortened the time to first relevant artifact from 42 to 14 min relative to unassisted examination (indicative estimates based on three examiner sessions; no inferential statistical test was performed). The study contributes a formalized, criteria-driven decision model for selecting between ISP and Chip-Off, which are experimentally validated thermal extraction profiles for eMMC, UFS, and PoP/RAM memory.