ABSTRACT
Edge-cloud video analytics systems capture video streams by edge cameras and send the video streams to the cloud for analytics to support applications like video surveillance, VR/AR, autonomous driving, etc. Video streams captured at the edge may contain sensitive objects, e.g., a human being. Existing studies propose adding noise to the intermediate video analytics results, encrypting video frames, etc. In this paper, we take an orthogonal approach where we remove, a.k.a. denaturing, the sensitive objects at the edge side before sending a video frame to the cloud.
Edge devices are highly resource-constrained, and the denaturing operation has non-trivial computation costs. More specifically, before denaturing, one needs to locate the sensitive objects by object detection; such object detection computation is resource intensive. In this paper, we propose EPC, an edge-cloud video analytics system that leverages a trajectory prediction model to locate sensitive objects in video frames. We formally analyze EPC and show that EPC can guarantee privacy. We evaluate EPC with two applications, person counting and vehicle detection. Evaluation results show that EPC can prevent privacy leakage under visual data attack with 95% video analytics accuracy and a 4x speedup compared to existing privacy control mechanisms.
- Alexandre Alahi, Kratarth Goel, Vignesh Ramanathan, et al. 2016. Social LSTM: Human Trajectory Prediction in Crowded Spaces. In Proc. of IEEE/CVF CVPR '16. Las Vegas, NV, USA.Google ScholarCross Ref
- Holger Caesar et al. 2020. nuScenes: A Multimodal Dataset for Autonomous Driving. In Proc. of IEEE/CFV CVPR '20. Virtual Event.Google Scholar
- S. Cass. 2019. Taking AI to the edge: Google's TPU now comes in a maker-friendly package. IEEE Spectrum 56, 5 (2019), 16--17.Google ScholarCross Ref
- Sandeep D'souza, Victor Bahl, et al. 2020. Amadeus: Scalable, Privacy-Preserving Live Video Analytics. arXiv:2011.05163 (2020).Google Scholar
- A. Elnagar. 2001. Prediction of moving objects in dynamic environments using Kalman filters. In Proc. of CIRA '01. Banff, Canada.Google ScholarCross Ref
- Junru Gu et al. 2021. DenseTNT: End-to-end Trajectory Prediction from Dense Goal Sets. In Proc. of IEEE/CVF ICCV '21. Virtual Event.Google Scholar
- Leal-Taixé et al. 2015. MOTChallenge 2015: Towards a Benchmark for Multi-Target Tracking. arXiv:1504.01942 (2015).Google Scholar
- Yuanqi Li et al. 2020. Reducto: On-Camera Filtering for Resource-Efficient Real-Time Video Analytics. In Proc. of ACM SIGCOMM '20. Virtual Event.Google Scholar
- Francesco Marchetti, Federico Becattini, Lorenzo Seidenari, and Alberto Del Bimbo. 2020. MANTRA: Memory Augmented Networks for Multiple Trajectory Prediction. In Proc. of IEEE/CFV CVPR '20. Virtual Event.Google ScholarCross Ref
- Fatemehsadat Mireshghallah, Mohammadkazem Taram, et al. 2020. Shredder: Learning Noise Distributions to Protect Inference Privacy. In Proc. of ACM ASPLOS '20. Lausanne, Switzerland.Google ScholarDigital Library
- Andrey Rudenko, Luigi Palmieri, et al. 2020. Human Motion Trajectory Prediction: A Survey. IJRR 39, 8 (2020), 895--935.Google ScholarDigital Library
- Liushuai Shi, Le Wang, Chengjiang Long, Sanping Zhou, et al. 2021. SGCN: Sparse Graph Convolution Network for Pedestrian Trajectory Prediction. In Proc. of IEEE/CVF CVPR '21. Virtual Event.Google Scholar
- Pieter Simoens et al. 2013. Scalable Crowd-sourcing of Video from Mobile Devices. In Proc. of ACM MobiSys '13. Taipei, Taiwan.Google Scholar
- Kimia Tajik, Akshith Gunasekaran, Rhea Dutta, et al. 2019. Balancing Image Privacy and Usability with Thumbnail-Preserving Encryption. In Proc. of NDSS Symposium '19. San Diego, CA, USA.Google ScholarCross Ref
- J. Wang et al. 2017. A scalable and privacy-aware IoT service for live video analytics. In Proc. of ACM MMSys '17. Taipei, Taiwan.Google ScholarDigital Library
- Hao Wu, Xuejin Tian, et al. 2021. PECAM: Privacy-Enhanced Video Streaming and Analytics via Securely-Reversible Transformation. In Proc. of ACM MobiCom '21. New Orleans, LA, USA.Google ScholarDigital Library
- Chenxin Xu, Weibo Mao, Wenjun Zhang, and Siheng Chen. 2022. Remember Intentions: Retrospective-Memory-based Trajectory Prediction. In Proc. of IEEE/CFV CVPR '22. New Orleans, LA, USA.Google ScholarCross Ref
- Guangsheng Zhang, Bo Liu, Tianqing Zhu, Andi Zhou, and Wanlei Zhou. 2022. Visual privacy attacks and defenses in deep learning: a survey. Artificial Intelligence Review 55, 6 (01 Aug 2022), 4347--4401.Google ScholarDigital Library
- Bolei Zhou et al. 2016. Learning Deep Features for Discriminative Localization. In Proc. of IEEE/CFV CVPR '16. Las Vegas, NV, USA.Google Scholar
Index Terms
- EPC: a video analytics system with efficient edge-side privacy control
Recommendations
Privacy leakage analysis in online social networks
Online Social Networks (OSNs) have become one of the major platforms for social interactions, such as building up relationship, sharing personal experiences, and providing other services. The wide adoption of OSNs raises privacy concerns due to personal ...
Strengthening EPC tags against cloning
WiSe '05: Proceedings of the 4th ACM workshop on Wireless securityThe EPC (Electronic Product Code) tag is a form of RFID (Radio-Frequency IDentification) device that is emerging as a successor to the printed barcode. Like barcodes, EPC tags emit static codes that serve to identify and track shipping containers and ...
Privacy-Enhancing of User's Behaviour Toward Privacy Settings in Social Networking Sites
CHI EA '16: Proceedings of the 2016 CHI Conference Extended Abstracts on Human Factors in Computing SystemsSocial Networking Sites (SNSs) are applications that allow users to create personal profiles to interact with friends or public and to share data such as photos and short videos. The amount of these personal disclosures has raised issues and concerns ...
Comments