Computing at the edge is increasingly important as Internet of Things (IoT) devices at the edge generate massive amounts of data and pose challenges in transporting all that data to the Cloud where they can be analyzed. On the other hand, harnessing the edge data is essential for offering cognitive applications, if the challenges, such as device capabilities, connectivity, and heterogeneity can be overcome. This paper proposes a novel three-tier architecture, called EdgeSphere, which harnesses resources of the edge devices, to analyze the data in situ at the edge. In contrast to the state-of-the-art cloud and mobile applications, EdgeSphere applications span across cloud, edge gateways, and edge devices. At its core, EdgeSphere builds on Apache Mesos to optimize resources usage and scheduling. EdgeSphere has been applied to practical scenarios and this paper describes the engineering challenges faced as well as innovative solutions.

Federated learning (FL) allows model training from local data collected by edge/mobile devices while preserving data privacy, which has wide applicability to image and vision applications. A challenge is that client devices in FL usually have much more limited computation and communication resources compared to servers in a data center. To overcome this challenge, we propose PruneFL —a novel FL approach with adaptive and distributed parameter pruning, which adapts the model size during FL to reduce both communication and computation overhead and minimize the overall training time, while maintaining a similar accuracy as the original model. PruneFL includes initial pruning at a selected client and further pruning as part of the FL process. The model size is adapted during this process, which includes maximizing the approximate empirical risk reduction divided by the time of one FL round. Our experiments with various datasets on edge devices (e.g., Raspberry Pi) show that: 1) we significantly reduce the training time compared to conventional FL and various other pruning-based methods and 2) the pruned model with automatically determined size converges to an accuracy that is very similar to the original model, and it is also a lottery ticket of the original model.

The Internet of Things (IoT) has become ubiquitous and greatly affected peoples’ daily lives. With the increasing development of IoT devices, the corresponding security issues are becoming more and more challenging. Such a severe security situation raises the following questions that need urgent attention: What are the primary security threats that IoT devices face currently? How do vendors and users deal with these threats? In this article, we aim to answer these critical questions through a large-scale systematic study. Specifically, we perform a ten-month-long empirical study on the vulnerability of 1,362,906 IoT devices varying from six types. The results show sufficient evidence that N-days vulnerability is seriously endangering the IoT devices: 385,060 (28.25 percent) devices suffer from at least one N-days vulnerability. Moreover, 2669 of these vulnerable devices may have been compromised by botnets. We further reveal the massive differences among five popular IoT search engines: Shodan [1], Censys [2], [3], Zoomeye [4], Fofa [5], and NTI [6]. To study whether vendors and users adopt defenses against the threats, we measure the security of MQTT [7] servers, and identify that 12740 (88 percent) MQTT servers have no password protection. Our analysis can serve as an important guideline for investigating the security of IoT devices, as well as advancing the development of a more secure environment for IoT systems.

Recent years have seen the increasing attention and popularity of federated learning (FL), a distributed learning framework for privacy and data security. However, by its fundamental design, federated learning is inherently vulnerable to model poisoning attacks: a malicious client may submit the local updates to influence the weights of the global model. Therefore, detecting malicious clients against model poisoning attacks in federated learning is useful in safety-critical tasks.However, existing methods either fail to analyze potential malicious data or are computationally restrictive. To overcome these weaknesses, we propose a robust federated learning method where the central server learns a supervised anomaly detector using adversarial data generated from a variety of state-of-the-art poisoning attacks. The key idea of this powerful anomaly detector lies in a comprehensive understanding of the benign update through distinguishing it from the diverse malicious ones. The anomaly detector would then be leveraged in the process of federated learning to automate the removal of malicious updates (even from unforeseen attacks).Through extensive experiments, we demonstrate its effectiveness against backdoor attacks, where the attackers inject adversarial triggers such that the global model will make incorrect predictions on the poisoned samples. We have verified that our method can achieve 99.0% detection AUC scores while enjoying longevity as the model converges. Our method has also shown significant advantages over existing robust federated learning methods in all settings. Furthermore, our method can be easily generalized to incorporate newly-developed poisoning attacks, thus accommodating ever-changing adversarial learning environments.

A flurry of fuzzing tools (fuzzers) have been proposed in the literature, aiming at detecting software vulnerabilities effectively and efficiently. To date, it is however still challenging to compare fuzzers due to the inconsistency of the benchmarks, performance metrics, and/or environments for evaluation, which buries the useful insights and thus impedes the discovery of promising fuzzing primitives. In this paper, we design and develop UNIFUZZ, an open-source and metrics-driven platform for assessing fuzzers in a comprehensive and quantitative manner. Specifically, UNIFUZZ to date has incorporated 35 usable fuzzers, a benchmark of 20 real-world programs, and six categories of performance metrics. We first systematically study the usability of existing fuzzers, find and fix a number of flaws, and integrate them into UNIFUZZ. Based on the study, we propose a collection of pragmatic performance metrics to evaluate fuzzers from six complementary perspectives. Using UNIFUZZ, we conduct in-depth evaluations of several prominent fuzzers including AFL [1], AFLFast [2], Angora [3], Honggfuzz [4], MOPT [5], QSYM [6], T-Fuzz [7] and VUzzer64 [8]. We find that none of them outperforms the others across all the target programs, and that using a single metric to assess the performance of a fuzzer may lead to unilateral conclusions, which demonstrates the significance of comprehensive metrics. Moreover, we identify and investigate previously overlooked factors that may significantly affect a fuzzer's performance, including instrumentation methods and crash analysis tools. Our empirical results show that they are critical to the evaluation of a fuzzer. We hope that our findings can shed light on reliable fuzzing evaluation, so that we can discover promising fuzzing primitives to effectively facilitate fuzzer designs in the future.

Edge devices use neural network models learnt on cloud to predict labels of its data records, which may lead to incorrect predictions especially for records that are different from the data involved in the training process, i.e., out-of-distribution (OOD) records. However, recent efforts in OOD detection either require the retraining of the model or assume the existence of a certain amount of OOD records, thus limiting their application in practice. In this work, we propose a novel OOD detection method (named as NeuralFP) without requiring any access to OOD records, which constructs non-linear fingerprints of neural network models memorizing the information of data observed during training. The key idea of NeuralFP is to exploit the difference in how the neural network model responds to data records in its training set versus data records that are anomalous. Specifically, NeuralFP builds autoencoders for each layer of the neural network model and then carefully analyzes the error distribution of the autocoders in reconstructing the training set to identify OOD records. Through extensive experiments on multiple real-world datasets, we show the effectiveness of NeuralFP in detecting OOD records as well as its advantages over previous approaches. Furthermore, we provide useful guidelines for parameter selection in the practical adoption of NeuralFP.

Supervised machine learning (ML) algorithms are aimed at maximizing classification performance under available energy and storage constraints. They try to map the training data to the corresponding labels while ensuring generalizability to unseen data. However, they do not integrate meaning-based relationships among labels in the decision process. On the other hand, natural language processing (NLP) algorithms emphasize the importance of semantic information. In this article, we synthesize the complementary advantages of supervised ML and NLP algorithms into one method that we refer to as SECRET (Semantically Enhanced Classification of REal-world Tasks). SECRET performs classifications by fusing the semantic information of the labels with the available data: it combines the feature space of the supervised algorithms with the semantic space of the NLP algorithms and predicts labels based on this joint space. Experimental results indicate that, compared to traditional supervised learning, SECRET achieves up to 14.0 percent accuracy and 13.1 percent F1 score improvements. Moreover, compared to ensemble methods, SECRET achieves up to 12.7 percent accuracy and 13.3 percent F1 score improvements. This points to a new research direction for supervised classification based on incorporation of semantic information.

Video cameras are among the most ubiquitous sensors in the Internet-of-Things. Video and audio applications, such as cross-camera activity detection, avatar extraction or language translation will, in the future, offload processing to an edge cluster of GPUs. Rim is a management system for such clusters that satisfies throughput and latency requirements of these applications, while enabling high cluster utilization. It uses coarse-grained knowledge of application structure to profile throughput of applications on resources, then uses these profiles to place applications on cluster nodes to achieve these goals. It dynamically adapts placement to load and failures. Experiments show that on maximal workloads on a testbed, Rim can satisfy requirements of all applications, but competing approaches designed for low-latency GPU execution cannot.

Many image and vision applications require a large amount of data for model training. Collecting all such data at a central location can be challenging due to data privacy and communication bandwidth restrictions. Federated learning is an effective way of training a machine learning model in a distributed manner from local data collected by client devices, which does not require exchanging the raw data among clients. A challenge is that among the large variety of data collected at each client, it is likely that only a subset is relevant for a learning task while the rest of data has a negative impact on model training. Therefore, before starting the learning process, it is important to select the subset of data that is relevant to the given federated learning task. In this paper, we propose a method for distributedly selecting relevant data, where we use a benchmark model trained on a small benchmark dataset that is task-specific, to evaluate the relevance of individual data samples at each client and select the data with sufficiently high relevance. Then, each client only uses the selected subset of its data in the federated learning process. The effectiveness of our proposed approach is evaluated on multiple real-world image datasets in a simulated system with a large number of clients, showing up to 25% improvement in model accuracy compared to training with all data.

We address energy-efficient placement of data and analytics components of composite analytics services on a wireless network to minimize execution-time energy consumption (computation and communication) subject to compute, storage and network resource constraints. We introduce an expressive analytics service hypergraph model for representing k-ary composability relationships (k ≥ 2) between various analytics and data components and leverage binary quadratic programming (BQP) to minimize the total energy consumption of a given placement of the analytics hypergraph nodes on the network subject to resource availability constraints. Then, after defining a potential energy functional Φ(·) to model the affinities of analytics components and network resources using analogs of attractive and repulsive forces in physics, we propose a decentralized Metropolis Monte Carlo (MMC) sampling method which seeks to minimize Φ by moving analytics and data on the network. Although Φ is non-convex, using a potential game formulation, we identify conditions under which the algorithm provably converges to a local minimum energy equilibrium placement configuration. Trace-based simulations of the placement of a deep-neural-network analytics service on a realistic wireless network show that for smaller problem instances our MMC algorithm yields placements with total energy within a small factor of BQP and more balanced workload distributions; for larger problems, it yields low-energy configurations while the BQP approach fails.

Large-scale distributed computing systems often contain thousands of distributed nodes (machines). Monitoring the conditions of these nodes is important for system management purposes, which, however, can be extremely resource demanding as this requires collecting local measurements of each individual node and constantly sending those measurements to a central controller. Meanwhile, it is often useful to forecast the future system conditions for various purposes such as resource planning/allocation and anomaly detection, but it is usually too resource-consuming to have one forecasting model running for each node, which may also neglect correlations in observed metrics across different nodes. In this paper, we propose a mechanism for collecting and forecasting the resource utilization of machines in a distributed computing system in a scalable manner. We present an algorithm that allows each local node to decide when to transmit its most recent measurement to the central node, so that the transmission frequency is kept below a given constraint value. Based on the measurements received from local nodes, the central node summarizes the received data into a small number of clusters. Since the cluster partitioning can change over time, we also present a method to capture the evolution of clusters and their centroids. As an effective way to reduce the amount of computation, time-series forecasting models are trained on the time-varying centroids of each cluster, to forecast the future resource utilizations of a group of local nodes. The effectiveness of our proposed approach is confirmed by extensive experiments using multiple real-world datasets.

In this paper, we present a novel incremental and decremental learning method for the least-squares support vector machine (LS-SVM). The goal is to adapt a pre-trained model to changes in the training dataset, without retraining the model on all the data, where the changes can include addition and deletion of data samples. We propose a provably exact method where the updated model is exactly the same as a model trained from scratch using the entire (updated) training dataset. Our proposed method only requires access to the updated data samples, the previous model parameters, and a unique, fixed-size matrix that quantifies the effect of the previous training dataset. Our approach can significantly reduce the storage requirement of model updating, preserve the privacy of unchanged training samples without loss of model accuracy, and enhance the computational efficiency. Experiments on real-world image dataset validate the effectiveness of our proposed method.

Capturing and managing the data needed to build effective machine learning models for custom IoT environments requires a great deal of effort. The amount of data generated from IoT devices is abundant, but tools to find datasets appropriate for the desired models are lacking. This paper presents a data capture system and data management catalog with solutions addressing the challenges of curating IoT data applied to purpose-built machine learning deployments.

Acoustic signals contain rich information of the environment. They can be used for detecting anomalous events such as in automated machine monitoring. In this demonstration, we present our acoustic anomaly detection system that captures acoustic signals and classifies them using machine learning techniques. Our system includes a server for sound management and model training, a mobile client for sound capturing and real-time classification, and a workbench that acts as a user interface. We will show the full operational pipeline of our system in this demonstration.

Smart speakers have been recently adopted and widely used in consumer homes, largely as a communication interface between human and machines. In addition, these speakers can be used to monitor sounds other than human voice, for example, to watch over elderly people living alone, and to notify if there are changes in their usual activities that may affect their health. In this paper, we focus on the sound classification using machine learning, which usually requires a lot of training data to achieve good accuracy. Our main contribution is a data augmentation technique that generates new sound by shuffling and mixing two existing sounds of the same class in the dataset. This technique creates new variations on both the temporal sequence and the density of the sound events. We show in DCASE 2018 Task 5 that the proposed data augmentation method with our proposed convolutional neural network (CNN) achieves an average of macroaveraged F1 score of 89.95% over 4 folds of the development dataset. This is a significant improvement from the baseline result of 84.50%. In addition, we also verify that our proposed data augmentation technique can improve the classification performance on the Urban Sound 8K dataset.

Machine learning is a promising technology for many modern applications. To train an effective machine learning model, a large amount of data is required. However, data may be created in different organizations and sharing data across organizational boundaries is difficult due to privacy concerns and communication bandwidth limitations. Data summarization is a technique for reducing the amount of data that needs to be shared, while preserving characteristics in the data that are useful for training machine learning models. In this paper, we present an overview of data summarization techniques, which can be useful for machine learning across organizational boundaries. We also discuss some possible applications related to these data summarization techniques and challenges for future research.

Large-scale distributed computing systems often contain thousands of distributed nodes (machines). Monitoring the conditions of these nodes is important for system management purposes, which, however, can be extremely resource demanding as this requires collecting local measurements of each individual node and constantly sending those measurements to a central controller. Meanwhile, it is often useful to forecast the future system conditions for various purposes such as resource planning/allocation and anomaly detection, but it is usually too resource-consuming to have one forecasting model running for each node, which may also neglect correlations in observed metrics across different nodes. In this paper, we propose a mechanism for collecting and forecasting the resource utilization of machines in a distributed computing system in a scalable manner. We present an algorithm that allows each local node to decide when to transmit its most recent measurement to the central node, so that the transmission frequency is kept below a given constraint value. Based on the measurements received from local nodes, the central node summarizes the received data into a small number of clusters. Since the cluster partitioning can change over time, we also present a method to capture the evolution of clusters and their centroids. As an effective way to reduce the amount of computation, time-series forecasting models are trained on the time-varying centroids of each cluster, to forecast the future resource utilizations of a group of local nodes. The effectiveness of our proposed approach is confirmed by extensive experiments using multiple real-world datasets.

The reporting on nocturnal sounds like cough and snore is not only relevant to follow the progress of respiratory diseases of patients but also to assess the quality of sleep of subjects. In this study, we discuss an audio analysis approach to count individual cough events and the duration of snore sounds in presence of air-conditioner noise through recordings of a smartphone and computationally efficient classifiers. A new audio data set of cough and snore sounds was acquired from 26 subjects. Energy threshold-based segmentation was applied to identify cough or snore events in the original low noise dataset. A k-nearest neighbor classifier was trained to merge cough phases belonging to the same cough event, to derive the proper ground-truth labeling. The original audio signal was augmented by the superposition of air-conditioner noise, with a signal-to-noise ratio of -40dB to 40dB, to enrich the training set of the binary classifier. Nine out of 40 mel-frequency cepstral coefficients in combination with the logarithm of energy from an entire cough or snore event were computed. Various classifiers, such as k-nearest neighbor (k-NN), rule-based classifier, decision tree, random forest, naive Bayes, and support vector machine were benchmarked against each other. The k-NN classifier with k=1 resulted in the highest F_1 scores of.85 and.88 in the binary classification task using generalized and personalized models, respectively, considering noise augmented samples. These results underline the potential of smartphones to objectively report on patient symptoms through audio recordings at night.

A wide array of military and commercial applications rely on the collection and processing of audio data. One approach to perform analytics and machine learning on such data is to upload and process them at a central server (e.g., cloud) which offers abundant processing resources and the ability to run sophisticated machine learning models and analytics on the audio data. This approach can be inefficient due to the low bandwidth and energy limitations of mobile devices as well as intermittent connectivity to a central collection point such as the cloud. It is also problematic as audio data are often highly sensitive and subject to privacy constraints. An alternative approach is to perform audio analytics at edge of the network where data is generated. The challenge in this approach is the requirement to perform analytics subject to resource constraints which limit performance and accuracy of predictive analytics. In this paper, we present a system for performing predictive analytics on audio data, where the training is executed on the cloud and the classification can be executed at the edge. We present the design principles and architecture of the system, and quantify the performance tradeoff of executing analytics at contemporary edge devices versus the cloud.