Search papers, labs, and topics across Lattice.
The paper introduces PPEDCRF, a privacy-preserving framework that protects location privacy in dashcam videos against background-based retrieval attacks. PPEDCRF employs a dynamic conditional random field (CRF) to identify and track location-sensitive background regions across video frames, and then injects calibrated perturbations based on a hierarchical sensitivity model. Experiments on driving datasets show that PPEDCRF significantly reduces location retrieval accuracy while maintaining competitive object detection and segmentation performance compared to baseline methods.
You can hide the location in your dashcam videos from prying eyes without sacrificing object detection accuracy by selectively perturbing only the most revealing background features.
Dashcam videos collected by autonomous or assisted-driving systems are increasingly shared for safety auditing and model improvement. Even when explicit GPS metadata are removed, an attacker can still infer the recording location by matching background visual cues (e.g., buildings and road layouts) against large-scale street-view imagery. This paper studies location-privacy leakage under a background-based retrieval attacker, and proposes PPEDCRF, a privacy-preserving enhanced dynamic conditional random field framework that injects calibrated perturbations only into inferred location-sensitive background regions while preserving foreground detection utility. PPEDCRF consists of three components: (i) a dynamic CRF that enforces temporal consistency to discover and track location sensitive regions across frames, (ii) a normalized control penalty (NCP) that allocates perturbation strength according to a hierarchical sensitivity model, and (iii) a utility-preserving noise injection module that minimizes interference to object detection and segmentation. Experiments on public driving datasets demonstrate that PPEDCRF significantly reduces location-retrieval attack success (e.g., Top-k retrieval accuracy) while maintaining competitive detection performance (e.g., mAP and segmentation metrics) compared with common baselines such as global noise, white-noise masking, and feature-based anonymization. The source code is in https://github.com/mabo1215/PPEDCRF.git
