Impact of Anonymization on Vehicle Detector Performance

Schnabel, Lukas; Matzka, Stephan; Stellmacher, Martin; Pätzold, Michael; Matthes, Elmar

doi:10.1109/ai4i46381.2019.00016

Cited by 7 publications

(3 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Early methods have been relying on obfuscation with a solid colored box, pixelization, random pixel shuffling, Gaussian blur and distortion. Schnabel et al [Schnabel et al 2019] evaluated several techniques and concluded that anonymization of personal data in the training set can impact the detection of vehicles at various degrees. However one would expect that anonymization would have a different impact on detection performance, depending how important the region we target is for feature learning.…”

Section: Implications Of Gdpr On Ai and Image Recognitionmentioning

confidence: 99%

Autonomous Vehicles: Data Protection and Ethical Considerations

Krontiris¹,

Grammenou²,

Terzidou³

et al. 2020

Computer Science in Cars Symposium

View full text Add to dashboard Cite

Section: Implications Of Gdpr On Ai and Image Recognitionmentioning

confidence: 99%

Autonomous Vehicles: Data Protection and Ethical Considerations

Krontiris¹,

Grammenou²,

Terzidou³

et al. 2020

Computer Science in Cars Symposium

View full text Add to dashboard Cite

“…The application of anonymization techniques inherently bears the risk of degrading the quality of the training dataset, potentially exerting an adverse impact on the efficacy of image recognition models [9]. Nonetheless, traditional anonymization methods remain the standard approach for concealing personal information within data, and their impact on model learning performance still requires more research to be conclusive.…”

Section: Introductionmentioning

confidence: 99%

Balancing Privacy and Accuracy: Exploring the Impact of Data Anonymization on Deep Learning Models in Computer Vision

Lee,

You

2024

IEEE Access

View full text Add to dashboard Cite

Computer vision has become indispensable in various applications, including autonomous driving, medical imaging, security and surveillance, robotics, and pattern recognition. In recent years, the quality of training data has emerged as a critical factor for ensuring effectiveness in real-world scenarios. However, the increasing stringency of privacy regulations in various regions necessitates careful handling of collected images for computer vision. Personal information within images is typically anonymized by applying anonymization patterns to remove it. Empirical findings underscore the significant influence of data quality on the training of deep learning models. Striking the right balance between privacy and recognition performance becomes paramount. Therefore, it is essential to understand how the anonymization of image datasets affects deep learning model performance. In this paper, we thoroughly analyze the effects of different anonymization techniques on the performance of deep learning-based models in computer vision tasks, with a particular emphasis on presenting a model-centric perspective, such as the type of deep learning model and the number of parameters. We aim to provide valuable insights and guidelines for selecting the optimal level of anonymization that strikes a balance between recognition accuracy and privacy protection.

show abstract

“…Using anonymisation technique; de-identification can be employed to remove personal information of the AC's users [238].…”

mentioning

confidence: 99%

Data Privacy Threat Modelling for Autonomous Systems: A Survey From the GDPR's Perspective

Azam

Michala

Ansari

et al. 2023

IEEE Trans. Big Data

View full text Add to dashboard Cite

Intelligence-based applications have been increasingly deployed in every field of life including smart homes, smart cities, healthcare services, and autonomous systems where personal data is collected across heterogeneous sources and processed using "black-box" algorithms in opaque centralised servers. As a consequence, preserving the data privacy and security of these applications is of utmost importance. In this respect, a modelling technique for identifying potential data privacy threats and specifying countermeasures to mitigate the related vulnerabilities in such AI-based systems plays a significant role in preserving and securing personal data. Various threat modelling techniques have been proposed such as STRIDE, LINDDUN, and PASTA but none of them is sufficient to model the data privacy threats in autonomous systems. Furthermore, they are not designed to model compliance with data protection legislation like the EU/UK General Data Protection Regulation (GDPR), which is fundamental to protecting data owners' privacy as well as to preventing personal data from potential privacy-related attacks. In this article, we survey the existing threat modelling techniques for data privacy threats in autonomous systems and then analyse such techniques from the viewpoint of GDPR compliance. Following the analysis, We employ STRIDE and LINDDUN in autonomous cars, a specific use-case of autonomous systems, to scrutinise the challenges and gaps of the existing techniques when modelling data privacy threats. Prospective research directions for refining data privacy threats & GDPR-compliance modelling techniques for autonomous systems are also presented.

show abstract

Impact of Anonymization on Vehicle Detector Performance

Cited by 7 publications

References 2 publications

Autonomous Vehicles: Data Protection and Ethical Considerations

Autonomous Vehicles: Data Protection and Ethical Considerations

Balancing Privacy and Accuracy: Exploring the Impact of Data Anonymization on Deep Learning Models in Computer Vision

Data Privacy Threat Modelling for Autonomous Systems: A Survey From the GDPR's Perspective

Contact Info

Product

Resources

About