Search

In this paper, we study scheduling of a queueing system with zero knowledge of instantaneous network conditions. We consider a one-hop single-server queueing system consisting of $K$ queues, each with time-varying and non-stationary arrival and service rates. Our scheduling approach builds on an innovative combination of adversarial bandit learning and Lyapunov drift minimization, without knowledge of the instantaneous network state (the arrival and service rates) of each queue. We then present two novel algorithms \texttt{SoftMW} (SoftMaxWeight) and \texttt{SSMW} (Sliding-window SoftMaxWeight), both capable of stabilizing systems that can be stablized by some (possibly unknown) sequence of randomized policies whose time-variation satisfies a mild condition. We further generalize our results to the setting where arrivals and departures only have bounded moments instead of being deterministically bounded and propose \texttt{SoftMW+} and \texttt{SSMW+} that are capable of stabilizing the system. As a building block of our new algorithms, we also extend the classical \texttt{EXP3.S} (Auer et al., 2002) algorithm for multi-armed bandits to handle unboundedly large feedback signals, which can be of independent interest.

An accountable algorithmic transparency report (ATR) should ideally investigate the (a) transparency of the underlying algorithm, and (b) fairness of the algorithmic decisions, and at the same time preserve data subjects' privacy. However, a provably formal study of the impact to data subjects' privacy caused by the utility of releasing an ATR (that investigates transparency and fairness), is yet to be addressed in the literature. The far-fetched benefit of such a study lies in the methodical characterization of privacy-utility trade-offs for release of ATRs in public, and their consequential application-specific impact on the dimensions of society, politics, and economics. In this paper, we first investigate and demonstrate potential privacy hazards brought on by the deployment of transparency and fairness measures in released ATRs. To preserve data subjects' privacy, we then propose a linear-time optimal-privacy scheme, built upon standard linear fractional programming (LFP) theory, for announcing ATRs, subject to constraints controlling the tolerance of privacy perturbation on the utility of transparency schemes. Subsequently, we quantify the privacy-utility trade-offs induced by our scheme, and analyze the impact of privacy perturbation on fairness measures in ATRs. To the best of our knowledge, this is the first analytical work that simultaneously addresses trade-offs between the triad of privacy, utility, and fairness, applicable to algorithmic transparency reports., 56 pages, 5 figures, accepted in ACM Journal of Data and Information Quality (JDIQ), Special Issue on Data Transparency

Deep neural networks training jobs and other iterative computations frequently include checkpoints where jobs can be canceled based on the current value of monitored metrics. While most of existing results focus on the performance of all jobs (both successfully completed and canceled), in this work we explore scheduling policies that improve the sojourn time of successful jobs, which are typically more valuable to the user. Our model assumes that each job has a known discrete size distribution (e.g., estimated from previous execution logs) where the largest size value indicates a successful completion, while other size values correspond to termination checkpoints. In the single-server case where all jobs are available for scheduling simultaneously, we prove that optimal schedules do not preempt jobs, even when preemption overhead is negligible. Based on this, we develop a scheduling policy that minimizes the sojourn time of successful jobs asymptotically, i.e., when the number of jobs grows to infinity. Through an extensive numerical study, we show that this policy performs better than existing alternatives even when the number of jobs is finite. For more realistic scenarios with multiple servers and dynamic jobs arrivals, we propose an online approach based on our single-server scheduling policy. Through an extensive simulation study, using real-world traces, we demonstrate that this online approach results in better average sojourn time for successful jobs as compared to existing techniques., We found that theorem III.4 is incorrect and we have a different theorem for the performance of the algorithm

With the growing workload of inference tasks on mobile devices, state-of-the-art neural architectures (NAs) are typically designed through Neural Architecture Search (NAS) to identify NAs with good tradeoffs between accuracy and efficiency (e.g., latency). Since measuring the latency of a huge set of candidate architectures during NAS is not scalable, approaches are needed for predicting end-to-end inference latency on mobile devices. Such predictions are challenging due to hardware heterogeneity, optimizations applied by ML frameworks, and the diversity of neural architectures. Motivated by these challenges, in this paper, we first quantitatively assess characteristics of neural architectures and mobile devices that have significant effects on inference latency. Based on this assessment, we propose a latency prediction framework which addresses these challenges by developing operation-wise latency predictors, under a variety of settings and a number of hardware devices, with multi-core CPUs and GPUs, achieving high accuracy in end-to-end latency prediction, as shown by our comprehensive evaluations. To illustrate that our approach does not require expensive data collection, we also show that accurate predictions can be achieved on real-world NAs using only small amounts of profiling data.

Federated learning allows multiple users to collaboratively train a shared classification model while preserving data privacy. This approach, where model updates are aggregated by a central server, was shown to be vulnerable to poisoning backdoor attacks: a malicious user can alter the shared model to arbitrarily classify specific inputs from a given class. In this paper, we analyze the effects of backdoor attacks on federated meta-learning, where users train a model that can be adapted to different sets of output classes using only a few examples. While the ability to adapt could, in principle, make federated learning frameworks more robust to backdoor attacks (when new training examples are benign), we find that even 1-shot~attacks can be very successful and persist after additional training. To address these vulnerabilities, we propose a defense mechanism inspired by matching networks, where the class of an input is predicted from the similarity of its features with a support set of labeled examples. By removing the decision logic from the model shared with the federation, success and persistence of backdoor attacks are greatly reduced., Comment: 13 pages, 19 figures, NeurIPS Workshop on Scalability, Privacy, and Security in Federated Learning (NeurIPS-SpicyFL), 2020

@TechReport{UCAM-CL-TR-926, author = {Pal, Ranjan and Psounis, Konstantinos and Kumar, Abhishek and Crowcroft, Jon and Hui, Pan and Golubchik, Leana and Kelly, John and Chatterjee, Aritra and Tarkoma, Sasu}, title = {{Are cyber-blackouts in service networks likely?: implications for cyber risk management}}, year = 2018, month = oct, url = {https://www.cl.cam.ac.uk/techreports/UCAM-CL-TR-926.pdf}, institution = {University of Cambridge, Computer Laboratory}, number = {UCAM-CL-TR-926} } Service liability interconnections among networked IT and IoT driven service organizations create potential channels for cascading service disruptions due to modern cybercrimes such as DDoS, APT, and ransomware attacks. The very recent Mirai DDoS and WannaCry ransomware attacks serve as famous examples of cyber-incidents that have caused catastrophic service disruptions worth billions of dollars across organizations around the globe. A natural question that arises in this context is “what is the likelihood of a cyber-blackout?”, where the latter term is defined as: “the probability that all (or a major subset of) organizations in a service chain become dysfunctional in a certain manner due to a cyber-attack at some or all points in the chain”. The answer to this question has major implications to risk management businesses such as cyber-insurance when it comes to designing policies by risk-averse insurers for providing coverage to clients in the aftermath of such catastrophic network events. In this paper, we investigate this question in general as a function of service chain networks and different loss distribution types. We show somewhat surprisingly (and discuss potential practical implications) that following a cyber-attack, the probability of a cyber-blackout and the increase in total service-related monetary losses across all organizations, due to the effect of (a) network interconnections, and (b) a wide range of loss distributions, are mostly very small, regardless of the network structure – the primary rationale behind the results being attributed to degrees of heterogeneity in wealth base among organizations, and Increasing Failure Rate (IFR) property of loss distributions.

The recent emergence of the small cloud (SC), both in concept and in practice, has been driven mainly by issues related to service cost and complexity of commercial cloud providers (e.g., Amazon) employing massive data centers. However, the resource inelasticity problem faced by the SCs due to their relatively scarce resources (e.g., virtual machines) might lead to a potential degradation of customer QoS and loss of revenue. A proposed solution to this problem recommends the sharing of resources between competing SCs to alleviate the resource inelasticity issues that might arise [1]. Based on this idea, a recent effort ([2]) proposed SC-Share, a performance-driven static market model for competitive small cloud environments that results in an efficient market equilibrium jointly optimizing customer QoS satisfaction and SC revenue generation. However, an important non-obvious question still remains to be answered, without which SC sharing markets may not be guaranteed to sustain in the long-run - is it still possible to achieve a stable market efficient state when the supply of SC resources is dynamic in nature and there is a variation of customer demand over time? In this paper, we address the problem of efficient market design for SC resource sharing in dynamic environments. We answer our previous question in the affirmative through the use of Arrow and Hurwicz's disequilibrium process [3], [4] in economics, and the gradient play technique in game theory that allows us to iteratively converge upon efficient and stable market equilibria

Small-scale clouds (SCs) often suffer from resource under-provisioning during peak demand, leading to inability to satisfy service level agreements (SLAs) and consequent loss of customers. One approach to address this problem is for a set of autonomous SCs to share resources among themselves in a cost-induced cooperative fashion, thereby increasing their individual capacities (when needed) without having to significantly invest in more resources. A central problem (in this context) is how to properly share resources (for a price) to achieve profitable service while maintaining customer SLAs. To address this problem, in this paper, we propose the SC-Share framework that utilizes two interacting models: (i) a stochastic performance model that estimates the achieved performance characteristics under given SLA requirements, and (ii) a market-based game-theoretic model that (as shown empirically) converges to efficient resource sharing decisions at market equilibrium. Our results include extensive evaluations that illustrate the utility of the proposed framework., Comment: To be published in ICDCS 2017

Despite the promising potential of network risk management services (e.g., cyber-insurance) to improve information security, their deployment is relatively scarce, primarily due to such service companies being unable to guarantee profitability. As a novel approach to making cyber-insurance services more viable, we explore a symbiotic relationship between security vendors (e.g., Symantec) capable of price differentiating their clients, and cyber-insurance agencies having possession of information related to the security investments of their clients. The goal of this relationship is to (i) allow security vendors to price differentiate their clients based on security investment information from insurance agencies, (ii) allow the vendors to make more profit than in homogeneous pricing settings, and (iii) subsequently transfer some of the extra profit to cyber-insurance agencies to make insurance services more viable. \noindent In this paper, we perform a theoretical study of a market for differentiated security product pricing, primarily with a view to ensuring that security vendors (SVs) make more profit in the differentiated pricing case as compared to the case of non-differentiated pricing. In order to practically realize such pricing markets, we propose novel and \emph{computationally efficient} consumer differentiated pricing mechanisms for SVs based on (i) the market structure, (ii) the communication network structure of SV consumers captured via a consumer's \emph{Bonacich centrality} in the network, and (iii) security investment amounts made by SV consumers., arXiv admin note: text overlap with arXiv:1101.5617 by other authors without attribution

Internet users such as individuals and organizations are subject to different types of epidemic risks such as worms, viruses, and botnets. To reduce the probability of risk, an Internet user generally invests in self-defense mechanisms like antivirus and antispam software. However, such software does not completely eliminate risk. Recent works have considered the problem of residual risk elimination by proposing the idea of cyber-insurance. In reality, an Internet user faces risks due to security attacks as well as risks due to non-security related failures (e.g., reliability faults in the form of hardware crash, buffer overflow, etc.) . These risk types are often indistinguishable by a naive user. However, a cyber-insurance agency would most likely insure risks only due to security attacks. In this case, it becomes a challenge for an Internet user to choose the right type of cyber-insurance contract as standard optimal contracts, i.e., contracts under security attacks only, might prove to be sub-optimal for himself. In this paper, we address the problem of analyzing cyber-insurance solutions when a user faces risks due to both, security as well as non-security related failures. We propose \emph{Aegis}, a novel cyber-insurance model in which the user accepts a fraction \emph{(strictly positive)} of loss recovery on himself and transfers rest of the loss recovery on the cyber-insurance agency. We mathematically show that given an option, Internet users would prefer Aegis contracts to traditional cyber-insurance contracts, under all premium types. This result firmly establishes the non-existence of traditional cyber-insurance markets when Aegis contracts are offered to users.

Internet users such as individuals and organizations are subject to different types of epidemic risks such as worms, viruses, spams, and botnets. To reduce the probability of risk, an Internet user generally invests in traditional security mechanisms like anti-virus and anti-spam software, sometimes also known as self-defense mechanisms. However, such software does not completely eliminate risk. Recent works have considered the problem of residual risk elimination by proposing the idea of cyber-insurance. In this regard, an important research problem is the analysis of optimal user self-defense investments and cyber-insurance contracts under the Internet environment. In this paper, we investigate two problems and their relationship: 1) analyzing optimal self-defense investments in the Internet, under optimal cyber-insurance coverage, where optimality is an insurer objective and 2) designing optimal cyber-insurance contracts for Internet users, where a contract is a (premium, coverage) pair.

Data, statistical models and inferential procedures permeate CENS deployments, from the four founding scientific application areas to the more recent urban sensing campaigns. This cross-center research breaks down into three classes of research: 1) General statistical models for embedded sensing, with specific applications to data quality and continuous sampling, 2) Significant CENS-designed and supported databases and repositories, and 3) Studies into the data lifecycle for embedded sensing systems.

Observing systems facilitate scientific studies by instrumenting the real world and collecting corresponding measurements, with the aim of detecting and tracking phenomena of interest. Our AMBROSia project focuses on a class of observing systems which are embedded into the environment, consist of stationary and mobile sensors, and react to collected observations by reconfiguring the system and adapting which observations are collected next. In this paper, we report on recent research directions and corresponding results in the context of AMBROSia.

Observing systems facilitate scientific studies by instrumenting the real world and collecting corresponding measurements, with the aim of detecting and tracking phenomena of interest. A wide range of critical environmental monitoring objectives in resource management, environmental protection, and public health all require distributed observing systems. The goal of such systems is to help scientists verify or falsify hypotheses with useful samples taken by the stationary and mobile units, as well as to analyze data autonomously to discover interesting trends or alarming conditions. In our project, we focus on a class of observing systems which are embedded into the environment, consist of stationary and mobile sensors, and react to collected observations by reconfiguring the system and adapting which observations are collected next. In this paper, we give an overview of our project in the context of a marine biology application.

Flow reshaping is used in time-sensitive networks (as in the context of IEEE TSN and IETF Detnet) in order to reduce burstiness inside the network and to support the computation of guaranteed latency bounds. This is performed using per-flow regulators (such as the Token Bucket Filter) or interleaved regulators (as with IEEE TSN Asynchronous Traffic Shaping). Both types of regulators are beneficial as they cancel the increase of burstiness due to multiplexing inside the network. It was demonstrated, by using network calculus, that they do not increase the worst-case latency. However, the properties of regulators were established assuming that time is perfect in all network nodes. In reality, nodes use local, imperfect clocks. Time-sensitive networks exist in two flavours: (1) in non-synchronized networks, local clocks run independently at every node and their deviations are not controlled and (2) in synchronized networks, the deviations of local clocks are kept within very small bounds using for example a synchronization protocol (such as PTP) or a satellite based geo-positioning system (such as GPS). We revisit the properties of regulators in both cases. In non-synchronized networks, we show that ignoring the timing inaccuracies can lead to network instability due to unbounded delay in per-flow or interleaved regulators. We propose and analyze two methods (rate and burst cascade, and asynchronous dual arrival-curve method) for avoiding this problem. In synchronized networks, we show that there is no instability with per-flow regulators but, surprisingly, interleaved regulators can lead to instability. To establish these results, we develop a new framework that captures industrial requirements on clocks in both non-synchronized and synchronized networks, and we develop a toolbox that extends network calculus to account for clock imperfections., ACM SIGMETRICS 2020 Boston, Massachusetts, USA June 8-12, 2020

Current practice in benchmarking commercial computer systems is to run a number of industry-standard benchmarks and to report performance numbers. The huge amount of machines and the large number of benchmarks for which performance numbers are published make it hard to observe clear performance trends though. In addition, these performance numbers for specific benchmarks do not provide insight into how applications of interest that are not part of the benchmark suite would perform on those machines. In this work we build a methodology for analyzing published commercial machine performance data sets. We apply statistical data analysis techniques, more in particular principal components analysis and cluster analysis, to reduce the amount of information to a manageable amount to facilitate its understanding. Visualizing SPEC CPU2000 performance numbers for 26 benchmarks and 1000+ machines in just a few graphs gives insight into how commercial machines compare against each other.In this work we build a methodology for analyzing published commercial machine performance data sets. We apply statistical data analysis techniques, more in particular principal components analysis and cluster analysis, to reduce the amount of information to a manageable amount to facilitate its understanding. Visualizing SPEC CPU2000 performance numbers for 26 benchmarks and 1000+ machines in just a few graphs gives insight into how commercial machines compare against each other. In addition, we provide a way of relating inherent program behavior to these performance numbers so that insights can be gained into how the observed performance trends relate to the behavioral characteristics of computer programs. This results in a methodology for the ubiquitous benchmarking problem of predicting performance of an application of interest based on its similarities with the benchmarks in a published industry-standard benchmark suite.