Search

We consider a system in which there exists two ISPs, one “big” Content Provider (CP), and a continuum of End-Users (EUs). One of the ISPs is neutral and the other is non-neutral. We consider that the CP can differentiate between ISPs by controlling the quality of the content she is offering on each one. We also consider that EUs have different levels of innate preferences for ISPs. We formulate a sequential game, and explicitly characterize all the possible Sub-game Perfect Nash Equilibria (SPNE) of the game. We prove that if an SPNE exists, it would be one of the five possible strategies each of which we explicitly characterize. We prove that when EUs have sufficiently low innate preferences for ISPs, a unique SPNE exists in which the neutral ISP would be driven out of the market. We also prove that when these preferences are sufficiently high, there exists a unique SPNE with a non-neutral outcome in which both ISPs are active. Numerical results reveal that the neutral ISP receives a lower payoff and the non-neutral ISP receives a higher payoff (most of the time) in a non-neutral scenario. However, we identify scenarios in which the non-neutral ISP loses payoff by adopting non-neutrality. We also show that a non-neutral regime yields a higher welfare for EUs in comparison to a neutral one if the market power of the non-neutral ISP is small, the sensitivity of EUs (respectively, the CP) to the quality is low (respectively, high), or a combinations of these factors. [ABSTRACT FROM AUTHOR]

With wide deployment of machine learning (ML)-based systems for a variety of applications including medical, military, automotive, genomic, multimedia, and social networking, there is great potential for damage from adversarial learning (AL) attacks. In this article, we provide a contemporary survey of AL, focused particularly on defenses against attacks on deep neural network classifiers. After introducing relevant terminology and the goals and range of possible knowledge of both attackers and defenders, we survey recent work on test-time evasion (TTE), data poisoning (DP), backdoor DP, and reverse engineering (RE) attacks and particularly defenses against the same. In so doing, we distinguish robust classification from anomaly detection (AD), unsupervised from supervised, and statistical hypothesis-based defenses from ones that do not have an explicit null (no attack) hypothesis. We also consider several scenarios for detecting backdoors. We provide a technical assessment for reviewed works, including identifying any issues/limitations, required hyperparameters, needed computational complexity, as well as the performance measures evaluated and the obtained quality. We then delve deeper, providing novel insights that challenge conventional AL wisdom and that target unresolved issues, including: robust classification versus AD as a defense strategy; the belief that attack success increases with attack strength, which ignores susceptibility to AD; small perturbations for TTE attacks: a fallacy or a requirement; validity of the universal assumption that a TTE attacker knows the ground-truth class for the example to be attacked; black, gray, or white-box attacks as the standard for defense evaluation; and susceptibility of query-based RE to an AD defense. We also discuss attacks on the privacy of training data. We then present benchmark comparisons of several defenses against TTE, RE, and backdoor DP attacks on images. The article concludes with a discussion of continuing research directions, including the supreme challenge of detecting attacks whose goal is not to alter classification decisions, but rather simply to embed, without detection, “fake news” or other false content. [ABSTRACT FROM AUTHOR]

Efficient and fair allocation of multiple types of resources is a crucial objective in a cloud/distributed computing cluster. Users may have diverse resource needs. Furthermore, diversity in server properties/capabilities may mean that only a subset of servers may be usable by a given user. In platforms with such heterogeneity, we identify important limitations in existing multi-resource fair allocation mechanisms, notably Dominant Resource Fairness and its follow-up work. To overcome such limitations, we propose a newserver-based approach; each server allocates resources by maximizing a per-serverutility function. We propose a specific class of utility functions which, when appropriately parameterized, adjusts the trade-off between efficiency and fairness, and captures a variety of fairness measures (such as our recently proposedPer-Server Dominant Share Fairness). We establish conditions for the proposed mechanism to satisfy certain properties that are generally deemed desirable, e.g., envy-freeness, sharing incentive, bottleneck fairness, and Pareto optimality. To implement our resource allocation mechanism, we develop an iterative algorithm which is shown to be globally convergent. Subsequently, we show how the proposed mechanism could be implemented in a distributed fashion. Finally, we carry out extensive trace-driven simulations to show the enhanced performance of our proposed mechanism over the existing ones. [ABSTRACT FROM AUTHOR]

Network function virtualization along with network service chaining and forwarding graphs envision a reduction in the respective cost that end users, service providers, and network operators are experiencing, while providing complete and high quality services. The allocation of these service chains in a pool of available cloud or data center resources is a challenging problem that can affect the overall performance of the offered network services. Furthermore, a number of challenges associated with the hardware capabilities and the available resources of the cloud infrastructure, along with possible collocation constraints between the components of the service chain, can exponentially increase the complexity of resource allocation. This paper examines how to improve the overall allocation performance of deploying service chains in a cloud environment satisfying server affinity, collocation, and latency constraints. The proposed method is inspired by a partitioning game, where the various components of a service chain are split in a set of partitions executed as virtual machines/containers in appropriate servers. We mathematically prove that a Nash equilibrium exists for our partitioning game corresponding to an optimal solution. By implementing the partitioning game as an iterative refinement process, we also experimentally validate that the proposed algorithm converges to the optimal solution. [ABSTRACT FROM PUBLISHER]