Video data has become central in many modern systems, from public surveillance to autonomous vehicles, but its widespread use raises serious concerns about exposing people’s identities and behaviors. This survey takes a structured look at how recent research has tried to address these concerns through video anonymization. We conduct a systematic review of the literature and organize existing work into a taxonomy that separates different anonymization strategies, the visual regions they target, and the assumptions they make about the video setting. By examining these categories, we highlight how current methods approach privacy protection and where they tend to fall short. A consistent pattern across the field is that most studies are designed for single-view videos, even though many real environments involve multiple synchronized cameras. Only a handful of works consider this multi-perspective setting, revealing a clear disconnect between research and real-world needs. We also review the datasets and evaluation metrics commonly used in anonymization studies and show that they rarely capture multi-view complexity, underscoring the need for more representative benchmarks. Finally, we discuss the utility goals that anonymization methods aim to preserve and outline key gaps that future work must address to support practical, multi-camera video applications.

Human motion is a behavioral biometric trait that can be used to identify individuals and infer private attributes such as medical conditions. This poses a serious threat to privacy as motion extraction from video and motion capture are increasingly used for a variety of applications, including mixed reality, robotics, medicine, and the quantified self. In order to protect the privacy of the tracked individuals, anonymization techniques that preserve the utility of the data are required. However, anonymizing motion data is a challenging task because there are many dependencies in motion sequences (such as physiological constraints) that, if ignored, make the anonymized motion sequence appear unnatural. In this paper, we propose Pantomime, a full-body anonymization technique for motion data, which uses foundation motion models to generate motion sequences that adhere to the dependencies in the data, thus keeping the utility of the anonymized data high. Our results show that Pantomime can maintain the naturalness of the motion sequences while reducing the identification accuracy to 10%.

The convergence of 6G and WiFi technologies promises unprecedented online connectivity, enabling immersive experiences such as the Metaverse, with a particular emphasis on VR gaming. Despite these advancements, they bring to light considerable privacy concerns, especially the ability of adversaries to deduce personal information from encrypted gaming network traffic. Focusing on VR headsets and Nintendo Switch consoles, this study explores the privacy implications within such network environments. By simulating the typical network conditions of online multiplayer games, we expose potential privacy breaches by adversaries from both WiFi and WAN, including mobile service providers. Classical machine learning algorithms can successfully classify games and gaming consoles with an accuracy exceeding 90%, further dissecting the network traffic to unveil distinct signatures and assess privacy risks.

Biometric data contains distinctive human traits such as facial features or gait patterns. The use of biometric data permits an individuation so exact that the data is utilized effectively in identification and authentication systems. But for this same reason, privacy protections become indispensably necessary. Privacy protection is extensively afforded by the technique of anonymization. Anonymization techniques protect sensitive personal data from biometrics by obfuscating or removing information that allows linking records to the generating individuals, to achieve high levels of anonymity. However, our understanding and possibility to develop effective anonymization relies, in equal parts, on the effectiveness of the methods employed to evaluate anonymization performance. In this paper, we assess the state-of-the-art methods used to evaluate the performance of anonymization techniques for facial images and for gait patterns. We demonstrate that the state-of-the-art evaluation methods have serious and frequent shortcomings. In particular, we find that the underlying assumptions of the state-of-the-art are quite unwarranted. State-of-the-art methods generally assume a difficult recognition scenario and thus a weak adversary. However, that assumption causes state-of-the-art evaluations to grossly overestimate the performance of the anonymization. Therefore, we propose a strong adversary which is aware of the anonymization in place. This adversary model implements an appropriate measure of anonymization performance. We improve the selection process for the evaluation dataset, and we reduce the numbers of identities contained in the dataset while ensuring that these identities remain easily distinguishable from one another. Our novel evaluation methodology surpasses the state-of-the-art because we measure worst-case performance and so deliver a highly reliable evaluation of biometric anonymization techniques.

Face images are a rich source of information that can be used to identify individuals and infer private information about them. To mitigate this privacy risk, anonymizations employ transformations on clear images to obfuscate sensitive information, all while retaining some utility. Albeit published with impressive claims, they sometimes are not evaluated with convincing methodology. Reversing anonymized images to resemble their real input --- and even be identified by face recognition approaches --- represents the strongest indicator for flawed anonymization. Some recent results indeed indicate that this is possible for some approaches. It is, however, not well understood, which approaches are reversible, and why. In this paper, we provide an exhaustive investigation in the phenomenon of face anonymization reversibility. Among other things, we find that 11 out of 15 tested face anonymizations are at least partially reversible and highlight how both reconstruction and inversion are the underlying processes that make reversal possible.

Gait recognition is the process of identifying humans from their bipedal locomotion such as walking or running. As such, gait data is privacy sensitive information and should be anonymized where possible. With the rise of higher quality gait recording techniques, such as depth cameras or motion capture suits, an increasing amount of detailed gait data is captured and processed. The introduction and rise of the Metaverse is an example of a potentially popular application scenario in which the gait of users is transferred onto digital avatars. As a first step towards developing effective anonymization techniques for high-quality gait data, we study different aspects of movement data to quantify their contribution to gait recognition. We first extract categories of features from the literature on human gait perception and then design experiments for each category to assess how much the information they contain contributes to recognition success. We evaluated the utility of gait perturbation by means of naturalness ratings in a user study. Our results show that gait anonymization will be challenging, as the data is highly redundant and inter-dependent.

A longstanding problem in computer privacy is that of data anonymization. One common approach is to present a query interface to analysts, and anonymize on a query-by-query basis. In practice, this approach often uses a standard database back end, and presents the query semantics of the database to the analyst.

Advance Driver Assistance Systems (ADAS) have gained huge attention in the last decades. One of the fundamental steps of the video processing chain is the detection of areas where the car can drive through, i.e. free-space. In this paper we present an approach for the detection of free-space which is based on image segmentation and classification of the obtained image segments. For the image segmentation step we use several state-of-the-art approaches. The classification is done by a random-forest classifier trained to label the image segments with one of three geometric classes (ground, sky, vertical) based on spatial, color and shape features. Segments labelled as ground are used to detect the free-space area in front of the car. Furthermore, a comparison of the results obtained by using different segmentation approaches is provided.