Isao Echizen

New Approach to Quantification of Privacy on Social Network Sites

Users may unintentionally reveal private information to the world on their blogs on social network sites (SNSs). Information hunters can exploit such disclosed sensitive information for the purpose of advertising, marketing, spamming, etc. We present a new metric to quantify privacy, based on probability and entropy theory. Simply by relying on the total leaked privacy value calculated with our metric, users can adjust the amount of information they reveal on SNSs. Previous studies focused on quantifying privacy for purposes of data mining and location finding. The privacy metric in this paper deals with unintentional leaks of information from SNSs. Our metric helps users of SNSs find how much privacy can be preserved after they have published sentences on their SNSs. It is simple, yet precise, which is proved through an experimental evaluation.

Isolation in Cloud Computing and Privacy-Enhancing Technologies

Cloud Computing lifts the borders between the access control domain of individuals’ and companies’ IT systems by processing their data within the application frameworks and virtualized runtime environments of Cloud service providers. A deployment of traditional security policies for enforcing confidentiality of Cloud users’ data would lead to a conflict with the availability of the Cloud’s software services: confidentiality of data would be assured but Cloud services would not be available for every user of a Cloud. This state-of-the-art contribution shows the analogy of the confidentiality of external data processing by Cloud services with mechanisms known and applied in privacy. Sustainability in Cloud is a matter of privacy, which in Cloud is called “isolation”.

Maintaining Picture Quality and Improving Robustness of Color Watermarking by Using Human Vision Models

Digital watermarks on pictures are more useful when they are better able to survive image processing operations and when they cause less degradation of picture quality. Random geometric distortion is one of the most difficult kinds of image processing for watermarks to survive because of the difficulty of synchronizing the expected watermark patterns to the watermarks embedded in pictures. This paper proposes three methods to improve a previous method that is not affected by this difficulty but that is insufficient in maintaining picture quality and treating other problems in surviving image processing. The first method determines the watermark strength in L*u*v* space, where human-perceived degradation of picture quality can be measured in terms of Euclidian distance, but embeds and detects watermarks in YUV space, where the detection is more reliable. The second method, based on the knowledge of image quantization, uses the messiness of color planes to hide watermarks. The third method reduces detection noises by preprocessing the watermarked image with orientation-sensitive image filtering, which is especially effective in picture portions where pixel values change drastically. Subjective evaluations have shown that these methods improved the picture quality of the previous method by 0.5 point of the mean evaluation score at the representative example case. On the other hand, the watermark strength of the previous method could be increased by 30% through 60% while keeping the same picture quality. Robustness to image processing has been evaluated for random geometric distortion, JPEG compression, Gaussian noise addition, and median filtering and it was clarified that these methods reduced the detection error ratio to 1/10 through 1/4. These methods can be applied not only to the previous method but also to other types of pixel-domain watermarking such as the Patchwork watermarking method and, with modification, to frequency-domain watermarking.

Framework Based on Privacy Policy Hiding for Preventing Unauthorized Face Image Processing

We put forward a framework to address a problem created by the rapidly spreading use of imaging devices and related to involuntarily or unintentionally photographed individuals: their pictures can accumulate additional meta information via face recognition systems and can be manually tagged via social networks and publishing platforms. With this framework a user can express his/her picture privacy policy in a machine readable format and (to some extent) automatically enforce it. An easily understandable flag system is used to define restrictions on picture usage and link ability. This policy is encoded in an unobtrusive way into wardrobe patterns and accessory designs with almost no impact on apparel appearance or social interaction.

An algorithm for k -anonymity-based fingerprinting

The anonymization of sensitive microdata (e.g. medical health records) is a widely-studied topic in the research community. A still unsolved problem is the limited informative value of anonymized microdata that often rules out further processing (e.g. statistical analysis). Thus, a tradeoff between anonymity and data precision has to be made, resulting in the release of partially anonymized microdata sets that still can contain sensitive information and have to be protected against unrestricted disclosure. Anonymization is often driven by the concept of k-anonymity that allows fine-grained control of the anonymization level. In this paper, we present an algorithm for creating unique fingerprints of microdata sets that were partially anonymized with k-anonymity techniques. We show that it is possible to create different versions of partially anonymized microdata sets that share very similar levels of anonymity and data precision, but still can be uniquely identified by a robust fingerprint that is based on the anonymization process.

Privacy Visor: Method for Preventing Face Image Detection by Using Differences in Human and Device Sensitivity

A method is proposed for preventing unauthorized face image revelation through unintentional capture of facial images. Methods such as covering the face and painting particular patterns on the face effectively prevent detection of facial images but hinder face-to-face communication. The proposed method overcomes this problem through the use of a device worn on the face that transmits near-infrared signals that are picked up by camera image sensors, which makes faces in captured images undetectable. The device is similar in appearance to a pair of eyeglasses, and the signals cannot be seen by the human eye, so face-to-face communication is not hindered. Testing of a prototype ”privacy visor” showed that captured facial images are sufficiently corrupted to prevent unauthorized face image revelation by face detection.

Tagging Disclosures of Personal Data to Third Parties to Preserve Privacy

Privacy in cloud computing is at the moment simply a promise to be kept by the software service providers. Users are neither able to control the disclosure of personal data to third parties nor to check if the software service providers have followed the agreed-upon privacy policy. Therefore, disclosure of the users‘ data to the software service providers of the cloud raises privacy risks. In this article, we show a privacy risk by the example of using electronic health records abroad. As a countermeasure by an ex post enforcement of privacy policies, we propose to observe disclosures of personal data to third parties by using data provenance history and digital watermarking.

An algorithm for collusion-resistant anonymization and fingerprinting of sensitive microdata

The collection, processing, and selling of personal data is an integral part of today’s electronic markets, either as means for operating business, or as an asset itself. However, the exchange of sensitive information between companies is limited by two major issues: Firstly, regulatory compliance with laws such as SOX requires anonymization of personal data prior to transmission to other parties. Secondly, transmission always implicates some loss of control over the data since further dissemination is possible without knowledge of the owner. In this paper, we extend an approach based on the utilization of k-anonymity that aims at solving both concerns in one single step - anonymization and fingerprinting of microdata such as database records. Furthermore, we develop criteria to achieve detectability of colluding attackers, as well as an anonymization strategy that resists combined efforts of colluding attackers on reducing the anonymization-level. Based on these results we propose an algorithm for the generation of collusion-resistant fingerprints for microdata.

Coupons as Monetary Incentives in Participatory Sensing

Participation of people is the most important factor in providing high quality of service in mobile sensing applications. In this paper we study coupons as incentives in order to stimulate users participation, especially in applications that rely on real-time data. We argue that coupons do not only function as incentives to increase user participation, but they can also direct more people to the targeted sensing area, increasing the overall utility of data for service providers. In this paper we study coupons in combination with multi-attributive auctions, which gives the additional advantage to service providers of not having to determine the coupon value that users would expect in exchange for their data. Instead users have to compete with each other to win the auction, choosing coupons of lower values. Even though the combination of coupons with multi-attributive auctions is very attractive for participatory sensing, we also highlight some of the problems coupons have and especially those connected with user privacy.

Preventing re-recording based on difference between sensory perceptions of humans and devices

We propose a method for preventing illegal re-recording of images and videos with digital camcorders. Conventional digital watermarking techniques involve embedding content ID into images and videos, which helps to identify the place and time where the actual content was recorded. However, digital watermarking technology does not control the illegal re-recording of digital content with camcorders. The proposed method prevents such re-recording by corrupting the recorded content with an invisible noise signal using CCD or CMOS devices during recording. In this way, the re-recorded content will be unusable. To validate the results of this proposed method, we developed a functional prototype system for preventing illegal re-recording and implemented it on a 100-inch cinema screen.