Edge Computing for Artificial Intelligence

Wang, Xiaofei; Han, Yiwen; Leung, Victor C. M.; Niyato, Dusit; Yan, Xueqiang; Chen, Xu

doi:10.1007/978-981-15-6186-3_7

Cited by 7 publications

(8 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This has resulted in hitherto inconceivable applications for edge AI, such as assisting radiologists in diagnosing diseases, driving automobiles on the highway, and even fertilizing plants [20]. Numerous experts and companies are discussing and adopting edge computing, which has its roots in the 1990s with the advent of content delivery networks that employ edge servers located near customers to provide web and gameplay videos [21]. Nearly every single industry today has tasks that might be better served by using edge AI [22].…”

Section: The Birth Of Edge Aimentioning

confidence: 99%

“…Another work [43] only focuses on edge computing and gives various future directions. Further, we have identified three relevant surveys [19,21,23] which are mainly focuses on Edge AI. Most of the survey articles on Edge AI have been published before 2021 except [23], but this [23] work is just an overview on Edge AI or a visionary work which offers only future directions in many relevant areas but not in-depth.…”

Section: Comparison With Related Surveysmentioning

confidence: 99%

See 1 more Smart Citation

Edge AI: A survey

Singh

Gill

2023

Internet of Things and Cyber-Physical Systems

135

View full text Add to dashboard Cite

Section: The Birth Of Edge Aimentioning

confidence: 99%

Section: Comparison With Related Surveysmentioning

confidence: 99%

Edge AI: A survey

Singh

Gill

2023

Internet of Things and Cyber-Physical Systems

135

View full text Add to dashboard Cite

“…The Fog Computing is an intermediate between Cloud Computing and Edge Computing with intermediate capacities of processing and storage allowing more important treatments than at the edge of the network [1]. MEC allows us to deploy services for mobile networks (Wi-Fi or 3GPP) [28]. It is obvious that these three levels of processing are brought to collaborate according to different modalities conditioned by the use cases such as cloud-edge, fog-edge, or cloud-fog-edge [29].…”

Section: Processing Levelmentioning

confidence: 99%

A New Edge Computing Architecture for IoT and Multimedia Data Management

2022

View full text Add to dashboard Cite

The Internet of Things and multimedia devices generate a tremendous amount of data. The transfer of this data to the cloud is a challenging problem because of the congestion at the network level, and therefore processing time could be too long when we use a pure cloud computing strategy. On the other hand, new applications requiring the processing of large amounts of data in real time have gradually emerged, such as virtual reality and augmented reality. These new applications have gradually won over users and developed a demand for near real-time interaction of their applications, which has completely called into question the way we process and store data. To address these two problems of congestion and computing time, edge architecture has emerged with the goal of processing data as close as possible to users, and to ensure privacy protection and responsiveness in real-time. With the continuous increase in computing power, amounts of memory and data storage at the level of smartphone and connected objects, it is now possible to process data as close as possible to sensors or directly on users devices. The coupling of these two types of processing as close as possible to the data and to the user opens up new perspectives in terms of services. In this paper, we present a new distributed edge architecture aiming to process and store Internet of Things and multimedia data close to the data producer, offering fast response time (closer to real time) in order to meet the demands of modern applications. To do this, the processing at the level of the producers of data collaborate with the processing ready for the users, establishing a new paradigm of short supply circuit for data transmission inspired of short supply chains in agriculture. The removing of unnecessary intermediaries between the producer and the consumer of the data improves efficiency. We named this new paradigm the Short Supply Circuit Internet of Things (SSCIoT).

show abstract

“…That is why we cannot directly apply existing white-box adversarial attacks. Although we can enumerate a list of rules for manual conversion, it is not general and easy to be out of date as there may be customized operations defined by developers and multiple deep learning frameworks [14]. Hence, the attack to the on-device model can be regarded as a grey-box attack and we propose a way to bypass that limitation so that a white-box attack can be feasible.…”

Section: Introductionmentioning

confidence: 99%

Smart App Attack: Hacking Deep Learning Models in Android Apps

Huang¹,

Chen²

2022

Preprint

View full text Add to dashboard Cite

On-device deep learning is rapidly gaining popularity in mobile applications. Compared to offloading deep learning from smartphones to the cloud, on-device deep learning enables offline model inference while preserving user privacy. However, such mechanisms inevitably store models on users' smartphones and may invite adversarial attacks as they are accessible to attackers. Due to the characteristic of the on-device model, most existing adversarial attacks cannot be directly applied for ondevice models. In this paper, we introduce a grey-box adversarial attack framework to hack on-device models by crafting highly similar binary classification models based on identified transfer learning approaches and pre-trained models from TensorFlow Hub. We evaluate the attack effectiveness and generality in terms of four different settings including pre-trained models, datasets, transfer learning approaches and adversarial attack algorithms. The results demonstrate that the proposed attacks remain effective regardless of different settings, and significantly outperform state-of-the-art baselines. We further conduct an empirical study on real-world deep learning mobile apps collected from Google Play. Among 53 apps adopting transfer learning, we find that 71.7% of them can be successfully attacked, which includes popular ones in medicine, automation, and finance categories with critical usage scenarios. The results call for the awareness and actions of deep learning mobile app developers to secure the on-device models. The code of this work is available at https://github.com/Jinxhy/SmartAppAttack.

show abstract

Edge Computing for Artificial Intelligence

Cited by 7 publications

References 7 publications

Edge AI: A survey

Edge AI: A survey

A New Edge Computing Architecture for IoT and Multimedia Data Management

Smart App Attack: Hacking Deep Learning Models in Android Apps

Contact Info

Product

Resources

About