Mohammad Mahdi Khalili

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.

Nowadays, rating systems play a crucial role in the attraction of customers to different services. However, as it is difficult to detect a fake rating, fraudulent users can potentially unfairly impact the rating’s aggregated score. This fraudulent behavior can negatively affect customers and businesses. To improve rating systems, in this paper, we take a novel mechanism-design approach to increase the cost of fake ratings while providing incentives for honest ratings. However, designing such a mechanism is a challenging task, as it is not possible to detect fake ratings since raters might rate a same service differently. Our proposed mechanism RewardRating is inspired by the stock market model in which users can invest in their ratings for services and receive a reward on the basis of future ratings. We leverage the fact that, if a service’s rating is affected by a fake rating, then the aggregated rating is biased toward the direction of the fake rating. First, we formally model the problem and discuss budget-balanced and incentive-compatibility specifications. Then, we suggest a profit-sharing scheme to cover the rating system’s requirements. Lastly, we analyze the performance of our proposed mechanism.

Personal information and other types of private data are valuable for both data owners and institutions interested in providing targeted and customized services that require analyzing such data. In this context, privacy is sometimes seen as a commodity: institutions (data buyers) pay individuals (or data sellers) in exchange for private data. In this study, we examine the problem of designing such data contracts, through which a buyer aims to minimize his payment to the sellers for a desired level of data quality, while the latter aim to obtain adequate compensation for giving up a certain amount of privacy. Specifically, we use the concept of differential privacy and examine a model of linear and nonlinear queries on private data. We show that conventional algorithms that introduce differential privacy via zero-mean noise fall short for the purpose of such transactions as they do not provide a sufficient degree of freedom for the contract designer to negotiate between the competing interests of the buyer and the sellers. Instead, we propose a biased randomized algorithm to generate differentially private output and show that this algorithm allows us to customize the privacy-accuracy tradeoff for each individual. We use a contract design approach to find the optimal contracts when using this biased algorithm to provide privacy and show that under this combination the buyer can achieve the same level of accuracy with a lower payment as compared to using the conventional, unbiased algorithms, while at the same time incurring lower privacy loss for the sellers.

We consider an InterDependent Security (IDS) game with networked agents where each agent chooses an effort/investment level for securing itself. The agents are interdependent in that the state of security of one agent depends not only on its own effort but also on that of the other agents, and each agent can benefit from other agents' security investment and effort. Since the agents are interdependent, they try to take advantage of others' security investment and effort and choose to under-invest, which leads to an inefficient Nash equilibrium (NE). While this has been analyzed extensively in the literature, in this study we take a different angle. Specifically, we consider the possibility of allowing agents to pool their resources, i.e., to both invest in themselves as well as in other agents. We show that the interaction of strategic and selfish agents under resource pooling (RP) improves the agents' effort level as well as their utility as compared to a scenario without resource pooling. We show that the social welfare (total utility) at the NE of the game with resource pooling is higher than the maximum social welfare attainable in a game without resource pooling but using an optimal incentive mechanism. Furthermore, we show that while voluntary participation in this latter scenario is not generally true, it is guaranteed under resource pooling.

Machine Learning (ML) models trained on data from multiple demographic groups can inherit representation disparity (Hashimoto et al., 2018) that may exist in the data: the model may be less favorable to groups contributing less to the training process; this in turn can degrade population retention in these groups over time, and exacerbate representation disparity in the long run. In this study, we seek to understand the interplay between ML decisions and the underlying group representation, how they evolve in a sequential framework, and how the use of fairness criteria plays a role in this process. We show that the representation disparity can easily worsen over time under a natural user dynamics (arrival and departure) model when decisions are made based on a commonly used objective and fairness criteria, resulting in some groups diminishing entirely from the sample pool in the long run. It highlights the fact that fairness criteria have to be defined while taking into consideration the impact of decisions on user dynamics. Toward this end, we explain how a proper fairness criterion can be selected based on a general user dynamics model.

Machine learning models developed from real-world data can inherit potential, preexisting bias in the dataset. When these models are used to inform decisions involving human beings, fairness concerns inevitably arise. Imposing certain fairness constraints in the training of models can be effective only if appropriate criteria are applied. However, a fairness criterion can be defined/assessed only when the interaction between the decisions and the underlying population is well understood. We introduce two feedback models describing how people react when receiving machine-aided decisions and illustrate that some commonly used fairness criteria can end with undesirable consequences while reinforcing discrimination.

Alternating direction method of multiplier (ADMM) is a powerful method to solve decentralized convex optimization problems. In distributed settings, each node performs computation with its local data and the local results are exchanged among neighboring nodes in an iterative fashion. During this iterative process the leakage of data privacy arises and can accumulate significantly over many iterations, making it difficult to balance the privacy-accuracy tradeoff. We propose Recycled ADMM (R-ADMM), where a linear approximation is applied to every even iteration, its solution directly calculated using only results from the previous, odd iteration. It turns out that under such a scheme, half of the updates incur no privacy loss and require much less computation compared to the conventional ADMM. Moreover, R-ADMM can be further modified (MR-ADMM) such that each node independently determines its own penalty parameter over iterations. We obtain a sufficient condition for the convergence of both algorithms and provide the privacy analysis based on objective perturbation. It can be shown that the privacy-accuracy tradeoff can be improved significantly compared with conventional ADMM.

This article highlights how cyber risk dependencies can be taken into consideration when underwriting cyber-insurance policies. This is done within the context of a base rate insurance policy framework, which is widely used in practice. Specifically, we show that there is an opportunity for an underwriter to better control the risk dependency and the risk spill-over, ultimately resulting in lower overall cyber risks across its portfolio. To do so, we consider a Service Provider (SP) and its customers as the interdependent insurer’s customers: a data breach suffered by the SP can cause business interruption to its customers. In underwriting both the SP and its customers, we show that the insurer can increase its profit by incentivizing the SP (through a discount on its premium) to invest more in security, thereby decreasing the chance of business interruption to the customers, and increasing social welfare. For comparison, we also consider a scenario where the insurer underwrites only the SP’s customers (but not the SP), and receives compensation from the SP’s insurance carrier when losses are attributed to the SP. We show how the insurer’s best strategy is to underwrite both the SP and its customers. We use an actual cyber-insurance policy and claims data to calibrate and substantiate our analytical findings.