Search

In Press / En Prensa The virtualization of radio access networks (vRAN) is the last milestone in the NFV revolution. However, the complexrelationship between computing and radio dynamics make vRAN resource control particularly daunting. We present vrAIn, a resourceorchestrator for vRANs based on deep reinforcement learning. First, we use an autoencoder to project high-dimensional context data(traffic and channel quality patterns) into a latent representation. Then, we use a deep deterministic policy gradient (DDPG) algorithmbased on an actor-critic neural network structure and a classifier to map contexts into resource control decisions.We have evaluated vrAIn experimentally, using an open-source LTE stack over different platforms, and via simulations over aproduction RAN. Our results show that: (i) vrAIn provides savings in computing capacity of up to 30% over CPU-agnostic methods;(ii) it improves the probability of meeting QoS targets by 25% over static policies; (iii) upon computing capacity under-provisioning,vrAIn improves throughput by 25% over state-of-the-art schemes; and (iv) it performs close to an optimal offline oracle. To ourknowledge, this is the first work that thoroughly studies the computational behavior of vRANs and the first approach to a model-freesolution that does not need to assume any particular platform or context. This work was partially supported by the European Commission through Grant No. 856709 (5Growth) and Grant No. 856950 (5G-TOURS); by Science Foundation Ireland (SFI) through Grant No. 17/CDA/4760; and AEI/FEDER through project AIM under Grant No. TEC2016-76465-C2-1-R. Furthermore, the work is closely related to the EU project DAEMON (Grant No. 101017109).

Although the number of solutions in combinatorial optimization problems (COPs) is finite, some problems grow exponentially and render exact approaches unfeasible. So, approximate methods, such as heuristics, are customary. Each heuristic usually specializes in specific kinds of problems. Hence, other approaches seek to merge their strengths. One of them is selection hyper-heuristics. However, they usually provide scarce information about their sensitivity. Illumination algorithms may fix this issue since they focus on exploration rather than exploitation while preserving the best solutions under different criteria. Still, literature falls short when merging both approaches, representing a knowledge gap. This work tests the feasibility of using an illumination algorithm, MAP-Elites (ME), for tuning a sequence-based selection hyper-heuristic model for Balanced Partition problems. We choose ME since other researchers have successfully applied it to a different COP. So, we may achieve a hyper-heuristic that represents the best combination of heuristics while simultaneously gaining intel on the performance of diverse alternatives. Our approach operates by creating a multi-dimensional map, where each design variable represents the application of a heuristic. Afterward, ME generates mutated sequences and tests them to determine if they represent a better-performing solution. We consider 1500 instances that include easy and hard instances, analyzed under different scenarios to test our approach. We also include limit instances that are neither easy nor hard. Our resulting data support the proposed approach, as it performs toe-to-toe with a synthetic oracle and may even outperform it. This represents an outstanding result, since a brute-force approach is needed to achieve such an oracle. So, merging ME and hyper-heuristics is a path worth pursuing. We also present how each parameter affects the model performance and identify the critical and virtually irrelevant ones. This serves as the groundwork for future works that focus on exploiting the most relevant parameters.

Because noisy, intermediate-scale quantum (NISQ) machines accumulate errors quickly, we need new approaches to designing NISQ-aware algorithms and assessing their performance. Algorithms with characteristics that appear less desirable under ideal circumstances, such as lower success probability, may in fact outperform their ideal counterparts on existing hardware. We propose an adaptation of Grover's algorithm, subdividing the phase flip into segments to replace a digital counter and complex phase flip decision logic. We applied this approach to obtaining the best solution of the MAX-CUT problem in sparse graphs, utilizing multi-control, Toffoli-like gates with residual phase shifts. We implemented this algorithm on IBM Q processors and succeeded in solving a 5-node MAX-CUT problem, demonstrating amplitude amplification on four qubits. This approach will be useful for a range of problems, and may shorten the time to reaching quantum advantage., Comment: 17 pages, 24 figures

Relational databases have fulfilled the role of a primary data repository for decades. With the development of modern technologies, the type of databases has come to the forefront which - unlike relational databases - have an unconventional structure which allows greater flexibility in data storage. Since the design and implementation of non-relational databases, which dates to the end of the 20th century, various types of non-relational databases have been proposed - the most popular non-relational databases include document, graphic, column and key-value databases. Although the flexibility of non-relational databases is an excellent feature in many respects, the freedom of the data structure and the freedom of defining data type cause considerable problems in number of database aspects. One of the situations in which the freedom of the data structure causes problems is the transformation of non-relational databases (the most common document database MongoDB or DynamoDB) to relational or other type of database. Undefined values often cause a termination of a running process and/or higher error rate in the transformation process. For these reasons, it is necessary to solve problems with undefined values either before or during the process of transformation and thus avoid application errors. To this end, we have created a module for solution of the problem of undefined values in the transformation of a non-relational database MongoDB to a relational database Oracle. The main purpose of this article is not to eliminate errors when starting the process, but to design and implement a module that can apply a dynamic solution based on input data and the selection of the target database when unknown values occur in the non-relational database MongoDB.

Currently, most of the databases are migrated to the cloud environment ensuring reliability, stability, and scalability. Companies do not need to have their own database managers and administrators. With the increase of the data amount, it is necessary to take loading and import methods into the account, to optimize the whole process and minimize time and resource demands. In comparison with the local server storage located physically in the company, there should be an emphasis on the communication channel reliability and process supervision. This paper summarizes the methods available to load data into the Oracle Cloud environment supplemented by the own technique definition. It is based on structure management and index extraction outside the main transaction by pointing to the Notofication_module. Thanks to that, loading and migration operations can be optimized reducing the costs.

Active learning strategies are used to acquire an enlarged labelled training set that allows the learner to achieve a better performance. To this end, unlabelled instances are carefully selected for labelling by an human expert in order to achieve the best performance with the smallest number of questions to this oracle. Several techniques exist to select the most informative samples within active learning strategies. However, the effectiveness of these methods when applied to problems with imbalanced classes was not studied before. In this paper, we focus on the improvement of instance selection strategies for active learning techniques in the presence of class imbalance. In an imbalanced setting, learning algorithms have difficulties to focus on the minority class due to its under-representation. Still, this is typically the class of interest for the end-user. This mismatch between the classes distribution and the goals of the end user is known as the class imbalance problem. In an active learning setting for class imbalance problems, this becomes an even more challenging issue. We propose two novel active learning algorithms, ALOD and ALOD-RE centered around selecting the most informative samples to be labelled, while considering the selection of possible minority class cases and the generation of synthetic minority class examples to improve the learner performance. To this end, our two proposed solutions combine: active learning procedures, resampling strategies, and anomaly detection methods. Through an extensive set of experiments we show that the incorporation of outlier detection and resampling techniques in the active learning procedure benefits the learners performance on imbalanced domains. The performance advantages of our proposed ALOD and ALOD-RE algorithms are clearly supported by our experimental results.

With the aim to reduce car accidents, autonomous driving attracted a lot of attentions these years. However, recently reported crashes indicate that this goal is far from being achieved. Hence, cost-effective testing of autonomous driving systems (ADSs) has become a prominent research topic. The classical model-based testing (MBT), i.e., generating test cases from test models followed by executing the test cases, is ineffective for testing ADSs, mainly because of the constant exposure to ever-changing operating environments, and uncertain internal behaviors due to employed AI techniques. Thus, MBT must be adaptive to guide test case generation based on test execution results in a step-wise manner. To this end, we propose a natural language and model-based approach, named LiveTCM, to automatically execute and generate test case specifications (TCSs) by interacting with an ADS under test and its environment. LiveTCM is evaluated with an open-source ADS and two test generation strategies: Deep Q-Network (DQN)-based and Random. Results show that LiveTCM with DQN can generate TCSs with 56 steps on average in 60 seconds, leading to 6.4 test oracle violations and covering 14 APIs per TCS on average.

An efficient generation of certain form of probability distributions is considered using both quantum and classical computing through suitable modifications to Grovers Algorithm. The suggested combination of the Oracle and the low-depth Local Diffuser is applied only once per circuit (irrespective of the problem size), followed by measurements for a number of shots and classical post-processing. This can facilitate Noisy Intermediate Scale Quantum (NISQ) implementation.

Today's global economy daily collects and produces enormous amounts of data. Making strategic and informed decisions based on data storage and analysis is the key to a successful business. That is why there is a growing demand in the market today for employees who have acquired appropriate competencies in databases during their education. Current higher education development policies emphasize the benefits of new teaching and learning approaches, such as student-centered learning. This paper describes teaching databases' experience using Oracle Academy program resources to create an innovative student-centered educational environment at the Polytechnic of Šibenik. Students' attitudes towards such a way of learning were also researched and presented.

Oracle-bone inscriptions (OBIs), the oldest hieroglyphs in China, were mainly carved on cattle scapulars and tortoise shells, as well as other animal bones. However, automatically extracting OBI characters is a rather complex task due to their differences in character size, orientation, alignment and noisy background. Conventional techniques like Laplacian operation, gradient-edge, or connected component, cannot obtain satisfying results. Therefore, in this paper, instance segmentation methods under deep convolutional neural network were exploited to extract OBIs automatically. More specifically, a SOTA weakly supervised instance segmentation model was introduced to solve this problem, considering that the pixel-level annotation is notoriously time-consuming compared to the bounding boxes annotation, which is extremely serious for the annotation of OBI images because annotators' lack of domain knowledge. The model was trained by 3228 oracle rubbing images and were tested on 312 ones. Results demonstrated that this method can provide a feasible way to automatically extract OBIs from rubbing images (as shown in Fig. 1).

The transmission and storage process for the power data in an intelligent grid has problems such as a single point of failure in the central node, low data credibility, and malicious manipulation or data theft. The characteristics of decentralization and tamper-proofing of blockchain and its distributed storage architecture can effectively solve malicious manipulation and the single point of failure. However, there are few safe and reliable data transmission methods for the significant number and various identities of users and the complex node types in the power blockchain. Thus, this paper proposes a collaborative control scheme between on-chain and off-chain power data based on the distributed oracle technology. By building a trusted on-chain transmission mechanism based on distributed oracles, the scheme solves the credibility problem of massive data transmission and interactive power data between smart contracts and off-chain physical devices safely and effectively. Analysis and discussion show that the proposed scheme can realize the collaborative control between on-chain and off-chain data efficiently, safely, and reliably.

Enhancing the trust of machine learning-based classifiers with large input spaces is a desirable goal; however, due to high labeling costs and limited resources, this is a challenging task. One solution is to use test input prioritization techniques that aim to identify and label only the most effective test instances. These prioritized test inputs can then be used with some popular testing techniques e.g., Metamorphic testing (MT) to test and uncover implementation bugs in computationally complex machine learning classifiers that suffer from the oracle problem. However, there are certain limitations involved with this approach, (i) using a small number of prioritized test inputs may not be enough to check the program correctness over a large variety of input scenarios, and (ii) traditional MT approaches become infeasible when the programs under test exhibit a non-deterministic behavior during training e.g., Neural Network-based classifiers. Therefore, instead of using MT for testing purposes, we propose a metamorphic relation to solve a data generation/labeling problem; that is, enhancing the test inputs effectiveness by extending the prioritized test set with new tests without incurring additional labeling costs. Further, we leverage the prioritized test inputs (both source and follow-up data sets) and propose a statistical hypothesis testing (for detection) and machine learning-based approach (for prediction) of faulty behavior in two other machine learning classifiers (Neural Network-based Intrusion Detection Systems). In our case, the problem is interesting in the sense that injected bugs represent the high accuracy producing mutated program versions that may be difficult to detect by a software developer. The results indicate that (i) the proposed statistical hypothesis testing is able to identify the induced buggy behavior, and (ii) Random Forest outperforms and achieves the best performance over SVM and k-NN algorithms.

Gradient Methods with Memory store at runtime part of the oracle information obtained at previous iterations. This model allows them to outperform classical gradient methods in many situations. Solving the inner problem associated with the model does incur an overhead but, for unconstrained problems where the objective has a Lipschitz gradient, this overhead has been shown to be minimal. In this work we propose an accelerated gradient method with memory applicable to composite problems. In our method the model overhead remains negligible even for constrained problems with non-differentiable objectives. Although the inner problem cannot be solved exactly, we propose a model and choose the starting point of the inner optimization scheme in a way that prevents the accumulation of errors as the algorithm progresses. Moreover, our method dynamically adjusts the convergence guarantees to exceed those of the Fast Gradient Method. The theoretical predictions are confirmed numerically on an image deblurring problem.

This paper presents ProMutator, a scalable security analysis framework that detects price oracle vulnerabilities before attacks occur. ProMutator’s core idea is to simulate price oracle attacks locally by mutating the data needed for price calculation. ProMutator analyzes existing transactions to reconstruct probable DeFi use patterns, thereby reducing the required simulation runs drastically. ProMutator does not require any examined contracts’ high-level source code. Additionally, ProMutator generates a report for each detected vulnerability to facilitate further investigation. In our evaluation, ProMutator successfully discovered five out of six known and 27 new price oracle vulnerabilities in DeFi protocols. Index Terms—DeFi, price oracle, price oracle attack

Structural analysis is the study of finding component functions for a given function. In this paper, we proceed with structural analysis of structures consisting of the S (nonlinear Substitution) layer and the A (Affine or linear) layer. Our main interest is the S 1 AS 2 structure with different substitution layers and large input/output sizes. The purpose of our structural analysis is to find the functionally equivalent oracle F* and its component functions for a given encryption oracle F(= S 2 ∘ A ∘ S 1 ). As a result, we can construct the decryption oracle F*−1 explicitly and break the one-wayness of the building blocks used in a White-box implementation. Our attack consists of two steps: S layer recovery using multiset properties and A layer recovery using differential properties. We present the attack algorithm for each step and estimate the time complexity. Finally, we discuss the applicability of S 1 AS 2 structural analysis in a White-box Cryptography environment.

Blockchain transactions compete for limited space in blockchain blocks. Miners prefer to include transactions with higher fees into new blocks. In the context of Ethereum, gas price price oracles predict fees such that transactions submitted at those fees make it into a block within a target delay. In practice, however, oracles are not accurate, which makes it difficult for distributed applications to operate predictable services in terms of price and performance.To understand oracle performance we define a new gas prediction accuracy metric. We demonstrate that oracles underprice transactions, causing them to miss the delay target, as well as overprice transactions, causing them to meet the delay target, but at a higher than necessary cost. We provide comparative analysis of five gas price oracles showing their relative accuracy, transaction accept rates, price stability, and discuss factors that influence oracle accuracy. We observe that the ETHGasStation oracle produces the most accurate and stable price predictions. For users that prefer to run their own oracle Web3.py provides comparable performance.

Smart Contracts cannot get external data directly due to the closure and determinacy of blockchain itself, limiting the extensibility of blockchain applications. Oracle is proposed to serve as a data feed to offer authenticated and deterministic external data to smart contracts. However, existing centralized oracles are efficient but vulnerable to targeted attacks and suffering from a single point of failure, while existing decentralized oracles are inefficient. The paper presents a trusted blockchain oracle based on a decentralized Trusted Execution Environment (TEE) network called Tora, which embraces both efficiency and availability. The key of Tora is a hybrid consensus mechanism with the confidential available group selection based on Proof-of-Availability (PoA). We have implemented a prototype of Tora on Ethereum with Intel Software Guard eXtensions (SGX) and evaluated it with a data-fetch use case on different measurements, including gas costs, off-chain execution time, group selection results and throughput. The results show that Tora provides considerable efficiency and scalability.

Buildings in residential and commercial sites consume close to 40 per cent of the world’s total energy produced and is growing at a steady pace. The need to lower the energy footprint is a matter of sustainability and active research for the smart building community. Recent trends in machine learning have led to significant work on occupancy detection in spaces by training isolated or ex-situ models, but with no reliability of performance on unknown spaces. Model applicability becomes questionable when the sensor value distribution is different from training data and in a real-life this is usually the case. Furthermore, analyzing a space on a floor-plan in silo obscures the holistic view of interactivity between building elements. In this paper, we propose the design of a generic building management system that auto-learns occupancy patterns and leverages spatial organization to deliver actionable insights on energy savings. We combine the building information with sensor signals into a Spatio-temporal activity graph, whose edges are dynamically updated based on occupancy. We introduce human-space interaction models to infer the human transmission capacity of each edge and compute an Eigenvalue score for all the spaces to derive automated checkpoints on space-wise appliance monitoring.

In this paper, we focus on the optimal consensus problem where a group of agents are required to collectively minimize a separable objective function. Moreover, we suppose each component is only equipped with an inexact first-order oracle. To solve this problem, we present a modified version of distributed subgradient method and investigate its performance under different information or step size conditions. We also give an example to verify the efficiency of our designs.

Testing computationally complex neural network-based applications (i.e. network intrusion detection systems) is a challenging task due to the absence of a test oracle. Metamorphic testing is a method to potentially solve the oracle problem when the correctness of individual output is difficult to determine. However, due to the stochastic nature of these applications, multiple runs with the same input can produce slightly different results; thus rendering traditional metamorphic testing technique inadequate. To address this problem, this paper proposes a statistical metamorphic testing technique to test neural network based Network Intrusion Detection Systems (N-IDSs) in a nondeterministic environment. We also performed mutation analysis to show the effectiveness of the proposed approach. The results show that the proposed method has a strong defect detection capability and is able to kill 100% implementation bugs in two neural network-based N-IDSs, and 66.66% in a neural network-based cancer prediction system.

We consider the problem of verifying complex learned controllers using distillation. In contrast to previous work, we require that the distilled model maintains behavioural fidelity with an oracle, defining the notion of non-divergent path length (NPL) as a metric. We demonstrate that current distillation approaches with proven accuracy bounds do not have high expected NPL and can be out-performed by naive behavioural cloning. We thus propose a distillation algorithm that typically gives greater expected NPL, improved sample efficiency, and more compact models. We prove properties of NPL maximization and demonstrate the performance of our algorithm on deep Q-network controllers for three standard learning environments that have been used in this context: Pong, CartPole and MountainCar.

Learning by interaction is the key to skill acquisition for most living organisms, which is formally called Reinforcement Learning (RL). RL is efficient in finding optimal policies for endowing complex systems with sophisticated behavior. All paradigms of RL utilize a system model for finding the optimal policy. Modeling dynamics can be done by formulating a mathematical model or system identification. Dynamic models are usually exposed to aleatoric and epistemic uncertainties that can divert the model from the one acquired and cause the RL algorithm to exhibit erroneous behavior. Accordingly, the RL process loses its generality because it is sensitive to operating conditions and changes in the model parameters. To address these problems, Intensive system identification for modeling purposes is needed for each system even if the model dynamics structure is the same, as the slight deviation in the model parameters can render the model useless in RL. The existence of an oracle that can adaptively predict the rest of the trajectory regardless of the uncertainties can help resolve the issue. The target of this work is to present a framework for facilitating the online system identification of different instances of the same dynamics class by learning a probability distribution of the dynamics conditioned on observed data with variational inference and show its reliability in robustly solving different instances of control problems without extra training in model-based RL with maximum sample efficiency.

International audience; We introduce in this paper a novel active learning algorithm for satellite image change detection. The proposed solution is interactive and based on a question and answer model, which asks an oracle (annotator) the most informative questions about the relevance of sampled satellite image pairs, and according to the oracle's responses, updates a decision function iteratively. We investigate a novel framework which models the probability that samples are relevant; this probability is obtained by minimizing an objective function capturing representativity, diversity and ambiguity. Only data with a high probability according to these criteria are selected and displayed to the oracle for further annotation. Extensive experiments on the task of satellite image change detection after natural hazards (namely tornadoes) show the relevance of the proposed method against the related work.

In this paper, we have conducted a thorough comparative analysis of the nested state between the two most popular hypervisors in the world today, which are Oracle's VirtualBox and VMware's workstation. The concept of Nested virtualization involves loading a virtual machine inside another virtual machine which could open a new dimension of multitasking and high-performance computing. In the present world scenario, this concept isn't as widely used as it has the potential to be. For our research, we set up the virtual machine on both VirtualBox and VMware workstation and measured the performance and did the same on the nested state. We then captured the slowdown of both the hypervisors and found out which virtual machine is better suited to perform the task of nested virtualization.

More recently, biometric systems have spread for modern security applications. Unfortunately, these systems have experienced several attempts of hacking. If biometric databases are compromised and stolen, biometrics in these databases will be lost forever. Consequently, there is an immediate need to introduce new upgradable biometric systems. The concept behind cancelable biometrics is to convert biometric data to alternative templates, which cannot be easily used by the impostor or intruder, and can be eliminated if breached. In this paper, the inverse filter is utilized in a cancelable face recognition system. In this system, masked biometric images are generated by blurring, noise addition and then inverse filtering. It is well-known in the image processing theory that inverse filtering leads to noise enhancement, which is an undesired effect in image restoration. In contrary, this effect will be desired in cancelable biometric systems. If the noise is magnified with an appropriate extent, it can mask the original biometrics leading to cancelable templates. This is the theory behind the proposed system. The proposed system is applied on the Olivetti and Oracle (ORL) dataset. Simulation results using evaluation metrics such as non-invertibility, unlinkability, visual inspection, False Positive Rate (FPR), False Negative Rate (FNR), Equal Error Rate (EER), Decidability, correlation coefficient, Area under the Receiver Operating Characteristic (AROC) curve demonstrate that the proposed system is resistant to intruders and hackers. Hence, it is efficient for several security applications.

Information is indispensable in modern society. People’s daily communication and work depend on information transmission. Unsafe storage or transmission of data may result in privacy and security problems. One way to attempt to prevent such issues is to use encryption algorithms to transform information into encrypted forms. Because the encryption steps of most encryption algorithms are complex, deciding the correctness of the encrypted output may take a long time in practice. This kind of problem is called the Test Oracle problem. In contrast to traditional software testing, Metamorphic Testing (MT) does not focus on the correctness of each individual output, but examines whether the inputs and outputs of multiple executions of a Program Under Test (PUT) satisfy necessary relations of the PUT, called metamorphic relations. This paper reports on an experience of applying MT to test three encryption algorithms — Data Encryption Standard (DES), Triple Data Encryption Standard (3DES), and Advanced Encryption Standard (AES).

Ethereum is a blockchain-based platform that pro-vides a global computational infrastructure to run smart contracts. In order to assign a cost to smart contract and transaction execution, the Ethereum Blockchain adopts a gas-based metering approach which is designed to motivate miners to operate the network and protect it against attacks. More precisely, miners receive fees from all transactions included in the mined block in addition to the mining reward. Hence, the higher the gas price in the transactions, the higher the fee paid to the miner will be, resulting in faster selection and execution of higher priced gas transactions. Therefore, an Ethereum transaction sender is exposed to the non-trivial task of having to choose an optimal gas price, as underpaying likely results in a transaction not being picked by miners, whereas overpaying leads to superfluous costs. This paper provides recommendation approach that proposes an appropriate gas price to users. More precisely, it investigates different approaches of forecasting algorithms applied for gas price predictions for the next block in Ethereum Blockchain. The gas price is predicted using the Prophet model and the deep learning models, Long-Short Term Memory (LSTM) and Gated Recurrent Unit (GRU). It also aims to compare these approaches with the most used gas price oracles. An evaluation of the obtained results show that the LSTM and GRU proposed models outperform Prophet model as well as the gas price oracle Geth. In this case, LSTM and GRU provide a low mean squared error (MSE) of 0,008 whereas Geth gives an MSE of 0.016 and Prophet gives an MSE of 0.014.

This paper proposes a new attack model where the attacker tries to reconstruct a combinational logic circuit without having full oracles access. This means due to limited access time to the product, the attacker has access to only a limited number of input-output (I/O) pairs and does not have any information about the design. The goal of the attacker is to reconstruct a circuit to be deployed with simulation or emulation, in order to act as an efficient surrogate to perform fast attacks.We propose Lorax, the first automated framework to reconstruct circuits from limited access, using tree-based machine learning (ML) models of different configurations. It features early estimation of accuracy of the reconstructed oracle using cross-validation, as well as approximation techniques for efficient synthesis of the learned logic. For cases that are difficult to learn, Lorax applies a special function matching phase utilizing an explanatory analysis of a tree-based ML model to identify bit importance. Our experiments show that with a training set of 6400 I/O pairs, Lorax can successfully approximate commonly-used functions from a range of sources, including arithmetic circuits, industrial designs, and computer vision problems, with an accuracy of 79–84% on average and near 100% for some arithmetic functions.

Since the popularization of social media, news has entered our lives digitally. While news is spreading broader and faster, fake news is becoming an increasingly popular topic. Fake news detection is therefore important in both social media and research areas. With artificial intelligence technology, software engineers have developed a lot of fake news detection systems. One of the biggest challenges for such systems is that they may face the oracle problem, which means that there may not be a way, or it may take too long time, to confirm the correctness of a specific output. Metamorphic Testing has been applied successfully to alleviate the oracle problem in many different areas, including in artificial intelligence. In this paper, we propose several metamorphic relations for fake news detection and report on experiments using metamorphic testing on fake news detection applications.

Block chain is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm. The blockchain is a closed system that is transparent to the outside world. No external logic can operate on chain data. The only node or interface that can add data to the blockchain is Oracles Machines. The oracle is a special node generated through a consensus mechanism. It is responsible for the ordering of on-chain requests, the simulation of the on-chain process, and the delivery of data for smart contracts. For example, the Chain Link node of Ethereum and the Orders node of Hyperledger. Although the blockchain is a decentralized structure, it is often a centralized and independent component outside the blockchain for the oracle. Its reliability, performance, and security are not protected by the blockchain system . How to protect the safety, reliability and efficiency of the oracle is a technical bottleneck of the blockchain. In response to the above-mentioned challenges, this paper studies the dynamic evolution configuration of the oracle in the case of a blockchain failure and the application software evolution algorithm based on the fault-tolerant oracle. The reflector technology is used in the research to monitor the dynamic changes of the oracle in real time. When a fault occurs, it immediately perceives the changes in the structure of the fault-tolerant blockchain to drive the evolution of the application software system and ensure the safe operation of the application system. Configure the fault-tolerant oracle Orderer environment based on the open source Fabric blockchain platform, and on this basis, experimentally verify the effectiveness of the application software evolution algorithm. Finally, a block chain application software system designed and implemented based on the above research results is introduced to further verify the effectiveness of the proposed algorithm from the perspective of practical application.

Internet of Things (IoT) is a key enabler of Industry 4.0 with networked devices providing sensor data to help manage, automate, streamline and optimize assets, operations and processes. In such industrial IoT settings, reliability and process experts spend a considerable amount of time in creating accurate ground-truth data to assist with the inferencing capabilities of Artificial Intelligence (AI) engines. This process can be time-consuming and sometimes inaccurate, depending on the complexity of data. Accurate expert annotated data is the foundation for many AI applications because data needs to be classified on several bases, for instance into ‘normal’, ‘abnormal’ or ‘pre-abnormal’ states. Such problem formulations can be appropriately addressed using Active Learning (AL) techniques. We propose an AL framework capable of handling two practical challenges: oracle uncertainty and quantification of model performance in the absence of ground truth. Consequently, the proposed approach addresses uncertainties within AL techniques by fusing information pertaining to expertise levels of the human annotators and their confidence levels corresponding to the annotation provided.

Automated bug report reproduction aims at assisting developers in efficiently replicating bugs reported by users on developers’ local machines. This activity involves the automation of reproduction trace generation and target bug verification (i.e. leveraging a bug oracle to check whether the triggered bug is the reported one). Recently, intensive studies have been conducted on automatically translating bug reports into executable reproduction traces. However, existing works all rely on crash stack trace to check if a triggered crash is same with the target crash, thus failing to reproduce a significant number of noncrash bugs. To address this problem, we propose an approach that utilizes an error screenshot in bug report as the bug oracle and applies a novel image comparison algorithm in comparing the oracle image with runtime screenshots. The proposed approach parses the oracle and runtime capture image pair using object detection techniques and then constructs a dummy hierarchy tree for structural difference comparison. In this way, it eliminates non-essential differences caused in most common bug reporting scenarios. We implemented our approach in a tool called SegStec. The empirical study upon 52 image pairs from 52 real-world Android bug reports indicates that SegStec outperforms the state-of-the-art pixel- level image comparison tool with an accuracy of 80.8% in target bug verification.

We are currently witnessing an impetuous development of blockchain technology in various fields like business, supply chains, or personal identity protection. Another technology that is becoming more widespread is the Internet of Things (IoT), which can be found in a lot of smart applications. In this paper, we aim to study how blockchain finds its utility in IoT applications. Finally, we analyze comparatively the reliability and availability of centralized and decentralized blockchain oracle mechanisms that provide data from IoT devices.

In this work, using physical features extracted from RF nonidealities in communicated EM signals, we show that radio frequency physical unclonable function (RF-PUF) performs much better compared to a solely convolutional neural network (CNN) based secure authentication method, ORACLE. For the static and quasi-static channels, respectively, we achieve 96% and 100% accuracy for RF-PUF compared to 87.13% and 98.6% accuracy for authentication using ORACLE. For the first time, RF-PUF has been applied for Wi-Fi devices to show that> 95% accuracy can be achieved for a wide range of transmitter and receiver separation from 2ft to 62ft both for the static and quasi-static channel, showing a peak of ~100% within 38ft range for the static case. The design space has been explored in detail. Finally, the concept of RF-PUF has been applied for clustering to detect safe-listed devices.

We establish fundamental limits of tracking a moving target over the unit cube under the framework of 20 questions with measurement-dependent noise. In this problem, there is an oracle who knows the instantaneous location of a target. Our task is to query the oracle as few times as possible to accurately estimate the trajectory of the moving target, whose initial location and velocity is unknown. We study the case where the oracle’s answer to each query is corrupted by random noise with query-dependent discrete distribution. In our formulation, the performance criterion is the resolution, which is defined as the maximal absolute value between the true location and estimated location at each discrete time during the searching process. We are interested in the minimal resolution of any non-adaptive searching procedure with a finite number of queries and derive approximations to this optimal resolution via the second-order asymptotic analysis.

End-to-end models that condition the output label sequence on all previously predicted labels have emerged as popular alternatives to conventional systems for automatic speech recognition (ASR). Since unique label histories correspond to distinct models states, such models are decoded using an approximate beam-search process which produces a tree of hypotheses. In this work, we study the influence of the amount of label context on the model's accuracy, and its impact on the efficiency of the decoding process. We find that we can limit the context of the recurrent neural network transducer (RNN-T) during training to just four previous word-piece labels, without degrading word error rate (WER) relative to the full-context baseline. Limiting context also provides opportunities to improve the efficiency of the beam-search process during decoding by removing redundant paths from the active beam, and instead retaining them in the final lattice. This path-merging scheme can also be applied when decoding the baseline full-context model through an approximation. Overall, we find that the proposed path-merging scheme is extremely effective allowing us to improve oracle WERs by up to 36% over the baseline, while simultaneously reducing the number of model evaluations by up to 5.3% without any degradation in WER.

Automatic speech recognition (ASR) systems often have difficulty recognizing long-tail entities such as contact names and local restaurant names, which usually do not occur, or occur infrequently, in the system’s training data. In this work, we present a method which uses learned text embeddings and nearest neighbor retrieval within a large database of entity embeddings to correct misrecognitions. Our text embeddings are produced by a neural network trained so that the embeddings of acoustically confusable phrases have low cosine distances. Given the embedding of the text of a potential entity misrecognition and a precomputed database containing entities and their corresponding embeddings, we use fast, scalable nearest neighbor retrieval algorithms to find candidate corrections within the database. The inserted candidates are then scored using a function of the original text’s cost in the lattice and the distance between the embedding of the original text and the embedding of the candidate correction. Using this lattice augmentation techique, we demonstrate a 46% reduction in word error rate (WER) and 46% reduction in oracle word error rate (OWER) on an evaluation set with popular film queries.

Automatic speech recognition (ASR) systems are highly sensitive to train-test domain mismatch. However, because transcription is often prohibitively expensive, it is important to be able to make use of available transcribed out-of-domain data. We address the problem of domain adaptation with semi-supervised training (SST). Contrary to work in in-domain SST, we find significant performance improvement even with just one hour of target-domain data—though, the selection of the data is critical. We show that minimum phone error rate is a good oracle measure for selection, and we approximate this measure by using the average phone confidence of an utterance. With larger domain shifts, we also find that deletions and low lexical diversity are a serious issue, which we address by incorporating phone rate into our selection metric. With our proposed selection criterion, we see up to 57% relative improvements over the out-of-domain baseline model. Furthermore, this selection method generalizes well, and matches or outperforms word-level confidence selection across six separate domain shift conditions.

As the performance of single-channel speech separation systems has improved, there has been a desire to move to more challenging conditions than the clean, near-field speech that initial systems were developed on. When training deep learning separation models, a need for ground truth leads to training on synthetic mixtures. As such, training in noisy conditions requires either using noise synthetically added to clean speech, preventing the use of in-domain data for a noisy-condition task, or training using mixtures of noisy speech, requiring the network to additionally separate the noise. We demonstrate the relative inseparability of noise and that this noisy speech paradigm leads to significant degradation of system performance. We also propose an SI-SDR-inspired training objective that tries to exploit the inseparability of noise to implicitly partition the signal and discount noise separation errors, enabling the training of better separation systems with noisy oracle sources., Accepted to ICASSP 2021

In the course of education reform, colleges and universities need to transform and upgrade the existing management mode, so as to meet the needs of the ever-expanding scale of students in colleges and universities. At the same time, it is also necessary to upgrade and transform the network teaching in colleges and universities. The education quality assessment of colleges and universities is very important for the management of colleges and universities. It can help the leaders of colleges and universities comprehensively and accurately grasp the teaching status quo, and improve the education quality of the whole college. The teaching quality directly reflects the teaching level of the whole university, so the evaluation of the whole university is very important. Teaching quality assessment is the overall assessment of teachers' teaching activities according to certain rules, including information statistics, students' learning conditions and students' feedback in class. Oracle databases are versatile and difficult to maintain and optimize. In the case of Oracle databases, the use of indexes, parallel control, SGA and PGA management, and the query speed of the database all interfere with the overall performance of the database.

Recently, semi-supervised domain adaptation (SSDA) approaches have shown impressive performance for the domain adaptation task. They effectively utilize few labeled target samples along with the unlabeled data to account for the distribution shift across the source and target domains. In this work, we make three-fold contributions, concentrating on the role of target samples and semantics for the SSDA task. First, we observe that choosing a few, but an equal number of labeled samples from each class in the target domain requires a significant amount of manual effort. To address this, we propose an active learning-based framework by modeling both the sample diversity and the classifier uncertainty. By utilizing k-means initialized cluster centers for picking a small pool of diverse unlabeled target samples, we compute a novel classifier adaptation uncertainty term to select the most effective samples from this pool, which are queried to obtain their true labels from an oracle. Second, we propose to weigh the hard target samples more, without explicitly using their predicted, possibly incorrect labels, which guides the adaptation process. Third, we note that irrespective of the domain shift, the semantics of the classes remain unchanged, so they can be effectively utilized for this task. We show that initializing the class-representations or prototypes with the class-semantics helps in bridging the domain gap significantly. These along with adversarially learnt entropy objective results in a novel framework, termed STar (Select TARgets), which sets a new state-of-the-art for the SSDA task.

Active learning selects the most critical instances and obtains their labels through interaction with an oracle. Selecting either informative or representative unlabeled instances may result in sampling bias or cluster dependency. In this paper, we propose a multi-standard optimization active learning (MSAL) algorithm that considers the informativeness, representativeness, and diversity of instances. Informativeness is measured by the soft-max predicted entropy, whereas representativeness is measured by the probability density function obtained by a non-parametric estimation. The multiplex of the two is used as an optimization objective to reduce model uncertainty and explore the distribution of unlabeled data. Diversity is measured by the difference between the selected critical instances. This is used as a constraint to prevent the selection of instances that are too similar. Learning experiments were performed with 12 datasets from various domains. The results of significance tests verify the effectiveness of MSAL and its superiority over state-of-the-art active learning algorithms.

Wireless sensor networks (WSN) are integral part of Industrial Internet of Things (IIOT), the said networks comprise of elements possessing low power processors. WSNs are used for gathering data in the monitoring region, using which vital information about the sensor and the monitoring region can be attained (placement of the sensor node is critical). Moreover, due to open nature of communication channel and resource constrained environment of nodes the privacy, integrity and confidentiality of the data becomes a big issue as we need to ensure that the said data is only accessed by a valid user in the IIOT environment. Many schemes presented various means to overcome the above issue. However, the existing works do have certain vulnerabilities which make its application in the WSNs constrained. The proposed work presents a scheme for achieving above goals in IIOT environment. Moreover, the proposed scheme presents security vulnerabilities and weaknesses in existing schemes and attempts to overcome these issues; the proposed work provides cryptanalysis of Xiong et al. scheme which is based on ECC, the direct consequence of the cryptanalysis directs to design a solution which can overcome the issues. The proposed work provides a robust hash based conditional privacy preserving authentication and probabilistic key exchange protocol which is lightweight and as a result, puts less computation overheads on the entities involved in the region. The proposed work is secure against many known attacks due to difficulty in guessing the credentials. The main motivation is to develop a lightweight scheme which can help exchange information with efficacy. The security of the proposed work is provided using both formal and informal security analysis where formal analysis comprises of AVISPA and Real-or-Random oracle model simulations whereas informal analysis depicts proofs as of how the proposed work withstands many known attacks. Thus, the proposed work is secure against many malicious attacks directing and championing its application in IIOT environment.