ACM Transactions on Privacy and Security最新文献

The 2019 Capital One data breach was one of the largest data breaches impacting the privacy and security of personal information of over a 100 million individuals. In most reports about a cyberattack, you will often hear that it succeeded because a single employee clicked on a link in a phishing email or forgot to patch some software, making it seem like an isolated, one-off, trivial problem involving maybe one person, committing a mistake or being negligent. But that is usually not the complete story. By ignoring the related managerial and organizational failures, you are leaving in place the conditions for the next breach. Using our Cybersafety analysis methodology, we identified control failures spanning control levels, going from rather technical issues up to top management, the Board of Directors, and Government regulators. In this analysis, we reconstruct the Capital One hierarchical cyber safety control structure, identify what parts failed and why, and provide recommendations for improvements. This work demonstrates how to discover the true causes of security failures in complex information systems and derive systematic cybersecurity improvements that likely apply to many other organizations. It also provides an approach that individuals can use to evaluate and better secure their organizations.

Many data analytics applications rely on temporal data, generated (and possibly acquired) sequentially for online analysis. How to release this type of data in a privacy-preserving manner is of great interest and more challenging than releasing one-time, static data. Because of the (potentially strong) temporal correlation within the data sequence, the overall privacy loss can accumulate significantly over time; an attacker with statistical knowledge of the correlation can be particularly hard to defend against. An idea that has been explored in the literature to mitigate this problem is to factor this correlation into the perturbation/noise mechanism. Existing work, however, either focuses on the offline setting (where perturbation is designed and introduced after the entire sequence has become available), or requires a priori information on the correlation in generating perturbation. In this study we propose an approach where the correlation is learned as the sequence is generated, and is used for estimating future data in the sequence. This estimate then drives the generation of the noisy released data. This method allows us to design better perturbation and is suitable for real-time operations. Using the notion of differential privacy, we show this approach achieves high accuracy with lower privacy loss compared to existing methods.

Anchor link prediction across social networks plays an important role in multiple social network analysis. Traditional methods rely heavily on user privacy information or high-quality network topology information. These methods are not suitable for multiple social networks analysis in real-life. Deep learning methods based on graph embedding are restricted by the impact of the active privacy protection policy of users on the graph structure. In this paper, we propose a novel method which neutralizes the impact of users’ evasion strategies. First, graph embedding with conditional estimation analysis is used to obtain a robust embedding vector space. Secondly, cross-network features space for supervised learning is constructed via the constraints of cross-network feature collisions. The combination of robustness enhancement and cross-network feature collisions constraints eliminate the impact of evasion strategies. Extensive experiments on large-scale real-life social networks demonstrate that the proposed method significantly outperforms the state-of-the-art methods in terms of precision, adaptability, and robustness for the scenarios with evasion strategies.

Current cloud deletion mechanisms fall short in meeting users’ various deletion needs. They assume all data is deleted the same way—data is temporally removed (or hidden) from users’ cloud accounts before being completely deleted. This assumption neglects users’ desire to have data completely deleted instantly or their preference to have it recoverable for a more extended period. To date, these preferences have not been explored. To address this gap, we conducted a participatory study with four groups of active cloud users (five subjects per group). We examined their deletion preferences and the information they require to aid deletion. In particular, we explored how users want to delete cloud data and identify what information about cloud deletion they consider essential, the time it should be made available to them, and the communication channel that should be used. We show that cloud deletion preferences are complex and multi-dimensional, varying between subjects and groups. Information about deletion should be within reach when needed, for instance, be part of deletion controls. Based on these findings, we discuss the implications of our study in improving the current deletion mechanism to accommodate these preferences.

We propose to identify compromised mobile devices from a network administrator’s point of view. Intuitively, inadvertent users (and thus their devices) who download apps through untrustworthy markets are often lured to install malicious apps through in-app advertisements or phishing. We thus hypothesize that devices sharing similar apps would have a similar likelihood of being compromised, resulting in an association between a compromised device and its apps. We propose to leverage such associations to identify unknown compromised devices using the guilt-by-association principle. Admittedly, such associations could be relatively weak as it is hard, if not impossible, for an app to automatically download and install other apps without explicit user initiation. We describe how we can magnify such associations by carefully choosing parameters when applying graph-based inferences. We empirically evaluate the effectiveness of our approach on real datasets provided by a major mobile service provider. Specifically, we show that our approach achieves nearly 98% AUC (area under the ROC curve) and further detects as many as 6 ~ 7 times of new compromised devices not covered by the ground truth by expanding the limited knowledge on known devices. We show that the newly detected devices indeed present undesirable behavior in terms of leaking private information and accessing risky IPs and domains. We further conduct in-depth analysis of the effectiveness of graph inferences to understand the unique structure of the associations between mobile devices and their apps, and its impact on graph inferences, based on which we propose how to choose key parameters.

Voice user interfaces and digital assistants are rapidly entering our lives and becoming singular touch points spanning our devices. These always-on services capture and transmit our audio data to powerful cloud services for further processing and subsequent actions. Our voices and raw audio signals collected through these devices contain a host of sensitive paralinguistic information that is transmitted to service providers regardless of deliberate or false triggers. As our emotional patterns and sensitive attributes like our identity, gender, well-being, are easily inferred using deep acoustic models, we encounter a new generation of privacy risks by using these services. One approach to mitigate the risk of paralinguistic-based privacy breaches is to exploit a combination of cloud-based processing with privacy-preserving, on-device paralinguistic information learning and filtering before transmitting voice data.

In this paper we introduce EDGY, a configurable, lightweight, disentangled representation learning framework that transforms and filters high-dimensional voice data to identify and contain sensitive attributes at the edge prior to offloading to the cloud. We evaluate EDGY’s on-device performance and explore optimization techniques, including model quantization and knowledge distillation, to enable private, accurate and efficient representation learning on resource-constrained devices. Our results show that EDGY runs in tens of milliseconds with 0.2% relative improvement in ‘zero-shot’ ABX score or minimal performance penalties of approximately 5.95% word error rate (WER) in learning linguistic representations from raw voice signals, using a CPU and a single-core ARM processor without specialized hardware.

The integrity of the content a user is exposed to when browsing the web relies on a plethora of non-web technologies and an infrastructure of interdependent hosts, communication technologies, and trust relations. Incidents like the Chinese Great Cannon or the MyEtherWallet attack make it painfully clear: the security of end users hinges on the security of the surrounding infrastructure: routing, DNS, content delivery, and the PKI. There are many competing, but isolated proposals to increase security, from the network up to the application layer. So far, researchers have focus on analyzing attacks and defenses on specific layers. We still lack an evaluation of how, given the status quo of the web, these proposals can be combined, how effective they are, and at what cost the increase of security comes. In this work, we propose a graph-based analysis based on Stackelberg planning that considers a rich attacker model and a multitude of proposals from IPsec to DNSSEC and SRI. Our threat model considers the security of billions of users against attackers ranging from small hacker groups to nation-state actors. Analyzing the infrastructure of the Top 5k Alexa domains, we discover that the security mechanisms currently deployed are ineffective and that some infrastructure providers have a comparable threat potential to nations. We find a considerable increase of security (up to 13% protected web visits) is possible at relatively modest cost, due to the effectiveness of mitigations at the application and transport layer, which dominate expensive infrastructure enhancements such as DNSSEC and IPsec.