Deep learning is at the heart of the current rise of artificial intelligence. In the field of computer vision, it has become the workhorse for applications ranging from self-driving cars to surveillance and security. Whereas, deep neural networks have demonstrated phenomenal success (often beyond human capabilities) in solving complex problems, recent studies show that they are vulnerable to adversarial attacks in the form of subtle perturbations to inputs that lead a model to predict incorrect outputs. For images, such perturbations are often too small to be perceptible, yet they completely fool the deep learning models. Adversarial attacks pose a serious threat to the success of deep learning in practice. This fact has recently led to a large influx of contributions in this direction. This paper presents the first comprehensive survey on adversarial attacks on deep learning in computer vision. We review the works that design adversarial attacks, analyze the existence of such attacks and propose defenses against them. To emphasize that adversarial attacks are possible in practical conditions, we separately review the contributions that evaluate adversarial attacks in the real-world scenarios. Finally, drawing on the reviewed literature, we provide a broader outlook of this research direction.

Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey

http://rboutaba.cs.uwaterloo.ca/Papers/Journals/2010/Mosharaf10.pdf

A survey of network virtualization

Emerging mega-trends (e.g., mobile, social, cloud, and big data) in information and communication technologies (ICT) are commanding new challenges to future Internet, for which ubiquitous accessibility, high bandwidth, and dynamic management are crucial. However, traditional approaches based on manual configuration of proprietary devices are cumbersome and error-prone, and they cannot fully utilize the capability of physical network infrastructure. Recently, software-defined networking (SDN) has been touted as one of the most promising solutions for future Internet. SDN is characterized by its two distinguished features, including decoupling the control plane from the data plane and providing programmability for network application development. As a result, SDN is positioned to provide more efficient configuration, better performance, and higher flexibility to accommodate innovative network designs. This paper surveys latest developments in this active research area of SDN. We first present a generally accepted definition for SDN with the aforementioned two characteristic features and potential benefits of SDN. We then dwell on its three-layer architecture, including an infrastructure layer, a control layer, and an application layer, and substantiate each layer with existing research efforts and its related research areas. We follow that with an overview of the de facto SDN implementation (i.e., OpenFlow). Finally, we conclude this survey paper with some suggested open research challenges.

A Survey on Software-Defined Networking

Recently network virtualization has been pushed forward by its proponents as a long-term solution to the gradual ossification problem faced by the existing Internet and proposed to be an integral part of the next-generation networking paradigm. By allowing multiple heterogeneous network architectures to cohabit on a shared physical substrate, network virtualization provides flexibility, promotes diversity, and promises security and increased manageability. However, many technical issues stand in the way of its successful realization. This article investigates the past and the state of the art in network virtualization along with the future challenges that must be addressed to realize a viable network virtualization environment.

/pdf/network-virtualization-state-of-the-art-and-research-4c2hab4lhv.pdf

Network virtualization: state of the art and research challenges

Since wireless network virtualization enables abstraction and sharing of infrastructure and radio spectrum resources, the overall expenses of wireless network deployment and operation can be reduced significantly. Moreover, wireless network virtualization can provide easier migration to newer products or technologies by isolating part of the network. Despite the potential vision of wireless network virtualization, several significant research challenges remain to be addressed before widespread deployment of wireless network virtualization, including isolation, control signaling, resource discovery and allocation, mobility management, network management and operation, and security as well as non-technical issues such as governance regulations, etc. In this paper, we provide a brief survey on some of the works that have already been done to achieve wireless network virtualization, and discuss some research issues and challenges. We identify several important aspects of wireless network virtualization: overview, motivations, framework, performance metrics, enabling technologies, and challenges. Finally, we explore some broader perspectives in realizing wireless network virtualization.

Wireless Network Virtualization: A Survey, Some Research Issues and Challenges

In this paper we present a programmable networking model that provides a common framework for understanding the state-of-the-art in programmable networks. A number of projects are reviewed and discussed against a set of programmable network characteristics. We believe that a number of important innovations are creating a paradigm shift in networking leading to higher levels of network programmability. These innovations include the separation between transmission hardware and control software, availability of open programmable network interfaces, accelerated virtualization of networking infrastructure, rapid creation and deployment of new network services and environments for resource partitioning and coexistence of multiple distinct network architectures. We present a simple qualitative comparison of the surveyed work and make a number of observations about the direction of the field.

/pdf/a-survey-of-programmable-networks-sqatw5xbte.pdf

A survey of programmable networks

Deep neural networks (DNNs) have achieved great success in solving a variety of machine learning (ML) problems, especially in the domain of image recognition. However, recent research showed that DNNs can be highly vulnerable to adversarially generated instances, which look seemingly normal to human observers, but completely confuse DNNs. These adversarial samples are crafted by adding small perturbations to normal, benign images. Such perturbations, while imperceptible to the human eye, are picked up by DNNs and cause them to misclassify the manipulated instances with high confidence. In this work, we explore and demonstrate how systematic JPEG compression can work as an effective pre-processing step in the classification pipeline to counter adversarial attacks and dramatically reduce their effects (e.g., Fast Gradient Sign Method, DeepFool). An important component of JPEG compression is its ability to remove high frequency signal components, inside square blocks of an image. Such an operation is equivalent to selective blurring of the image, helping remove additive perturbations. Further, we propose an ensemble-based technique that can be constructed quickly from a given well-performing DNN, and empirically show how such an ensemble that leverages JPEG compression can protect a model from multiple types of adversarial attacks, without requiring knowledge about the model.

Keeping the Bad Guys Out: Protecting and Vaccinating Deep Learning with JPEG Compression

The rapidly growing body of research in adversarial machine learning has demonstrated that deep neural networks (DNNs) are highly vulnerable to adversarially generated images. This underscores the urgent need for practical defense techniques that can be readily deployed to combat attacks in real-time. Observing that many attack strategies aim to perturb image pixels in ways that are visually imperceptible, we place JPEG compression at the core of our proposed SHIELD defense framework, utilizing its capability to effectively "compress away" such pixel manipulation. To immunize a DNN model from artifacts introduced by compression, SHIELD "vaccinates" the model by retraining it with compressed images, where different compression levels are applied to generate multiple vaccinated models that are ultimately used together in an ensemble defense. On top of that, SHIELD adds an additional layer of protection by employing randomization at test time that compresses different regions of an image using random compression levels, making it harder for an adversary to estimate the transformation performed. This novel combination of vaccination, ensembling, and randomization makes SHIELD a fortified multi-pronged defense. We conducted extensive, large-scale experiments using the ImageNet dataset, and show that our approaches eliminate up to 98% of gray-box attacks delivered by strong adversarial techniques such as Carlini-Wagner's L2 attack and DeepFool. Our approaches are fast and work without requiring knowledge about the model.

https://dl.acm.org/doi/pdf/10.1145/3219819.3219910

SHIELD: Fast, Practical Defense and Vaccination for Deep Learning using JPEG Compression

Existing mobile systems (e.g., mobile IP, mobile ATM, and third-generation cellular systems) lack the intrinsic architectural flexibility to deal with the complexity of supporting adaptive mobile applications in wireless and mobile environments. We believe that there is a need to develop alternative network architectures from the existing ones to deal with the tremendous demands placed on underlying mobile signaling, adaptation management, and wireless transport systems in support of new mobile services (e.g., interactive multimedia and Web access). We present the design, implementation, and evaluation of mobiware, a mobile middleware toolkit that enables adaptive mobile services to dynamically exploit the intrinsic scalable properties of mobile multimedia applications in response to time-varying mobile network conditions. The mobiware toolkit is software-intensive and is built on CORBA and Java distributed object technology. Based on an open programmable paradigm developed by the COMET Group, mobiware runs on mobile devices, wireless access points, and mobile-capable switch/routers providing a set of open programmable interfaces and algorithms for adaptive mobile networking.

https://cs.dartmouth.edu/~campbell/papers/pcs1998.pdf

The mobiware toolkit: programmable support for adaptive mobile networking

A method and apparatus for two-stage packet classification, the two-stage packet classification scheme including a first stage and a second stage. In the first classification stage, a packet is classified on the basis of the packet's network path. In the second stage of classification, the packet is classified on the basis of one or more transport (or other) fields of the packet. Also disclosed are embodiments of most specific filter matching and transport level sharing, and either one or both of these techniques may be implemented in the two-stage classification method.

Michael E. Kounavis

Papers

A survey of programmable networks

Keeping the Bad Guys Out: Protecting and Vaccinating Deep Learning with JPEG Compression

SHIELD: Fast, Practical Defense and Vaccination for Deep Learning using JPEG Compression

The mobiware toolkit: programmable support for adaptive mobile networking

Method and apparatus for two-stage packet classification using most specific filter matching and transport level sharing