Session GI-00

Session#0 - Opening Session with Invited Keynote

8:00 AM — 8:30 AM EDT
May 20 Sat, 8:00 AM — 8:30 AM EDT

Invited Keynote: Economic Mechanisms in Networking to Enable Innovation

Tilman Wolf (University of Massachusetts Amherst, USA)

This talk does not have an abstract.
Speaker Tilman Wolf (University of Massachusetts Amherst, USA)
Tilman Wolf is senior vice provost for Academic Affairs, Interim Department Head of Biomedical Engineering, and professor of Electrical and Computer Engineering, University of Massachusetts Amherst. He is a co-author of the book Architecture of Network Systems and has published extensively in peer-reviewed journals and conferences. His research has been supported by grants from NSF, DARPA, and industry. For additional info, please refer to his website.

Session Chair

Paolo Bellavista

Session GI-01

Session#1 - Applying Machine Learning to Next Generation Internet

8:30 AM — 9:30 AM EDT
May 20 Sat, 8:30 AM — 9:30 AM EDT

Multi Agent Reinforcement Learning based local routing strategy to reduce end-to-end delays in Segment Routing networks

Antonio Cianfrani (University of Rome Sapienza, Italy); Davide Aureli (Uniroma1, Italy); Marco Listanti (University of Rome "La Sapienza", Italy); Marco Polverini (University "La Sapienza" Roma, Italy)

In this paper we propose a framework based on Deep Reinforcement Learning to proactively and autonomously take under control links loads in Segment Routing (SR) networks. The main idea is to monitor local link loads and, in case of anomalous situation, to execute local routing changes at milliseconds timescale. The solution proposed is based on a Multi Agent Reinforcement Learning (MARL) approach: a subset of nodes is equipped with a local agent, powered by a Deep Q-Network (DQN) algorithm, referred to as Rv6 rerouting for Local In-network Link Load Control (SR-LILLC). The main feature of SR-LILLC is to train the agents in a collaborative way, by defining a "shared" reward function, while working in an independent way during the operating phase. Moreover, the re-routing operation is performed in a transparent way for other network devices, without involving the centralized control plane, by exploiting the source routing feature of the SR. The performance evaluation conducted over real data sets shows that SR-LILLC is able to reduce the load on agents links without increasing the maximum link utilization of the network; moreover, the overall network performance are improved in terms of end-to-end delays.
Speaker Antonio Cianfrani (University of Rome Sapienza)

Antonio Cianfrani received the "Laurea" degree, magna cum laude, in Telecommunications Engineering in 2004 and the Ph.D degree in Information and Communication Engineering in 2008 from University of Rome "La Sapienza". Since March 2019 he is an Associate Professor at the Department of Information, Electronic and Telecommunications engineering (DIET) of University of Rome “La Sapienza”. His main research interests include IP Routing protocols, Traffic Engineering, Segment routing and cloud networking.

Few Shot Learning Approaches for Classifying Rare Mobile-App Encrypted Traffic Samples

Giampaolo Bovenzi (University of Napoli Federico II, Italy); Davide Di Monda (IMT School for Advanced Studies Lucca & University of Napoli Federico II, Italy); Antonio Montieri (University of Napoli Federico II, Italy); Valerio Persico (University of Napoli Federico II, Italy); Antonio PescapÈ (University of Napoli Federico II, Italy)

Deep Learning (DL) is effective for classifying encrypted network traffic. However, it requires large amounts of labeled data to feed typical data-hungry training processes. Unfortunately, collecting and labeling rich network-traffic datasets is a complex and costly procedure not always affordable in practice, possibly hindering DL solutions. Few Shot Learning (FSL) aims at tackling this shortcoming, providing means to leverage non-few knowledge to support classification tasks related to traffic with few labeled data available. Although FSL has been largely investigated in other domains (e.g. computer vision), it has been only preliminarily adopted for the classification of encrypted traffic. In this work, we provide a first attempt in adopting FSL for classifying mobile-app encrypted traffic. Specifically, we consider the two most popular FSL paradigms: meta learning (learn to learn) and transfer learning (knowledge transfer from related tasks). We consider a number of variants for each (namely MatchingNet, ProtoNet, RelationNet, MetaOptNet, fo-MAML, ANIL, Fine-Tuning, and Freezing) and provide an empirical assessment of these approaches when adopted for mobile-app traffic classification considering the Mirage-2019 dataset as a test bench. Results show that FSL in mobile-app traffic classification is feasible, reaching satisfactory results (up to 80% F1-score), but leaving room for improvement.
Speaker Davide Di Monda (IMT School for Advanced Studies Lucca & University of Napoli Federico II, Italy)

Davide Di Monda is a PhD student at IMT School for Advanced Studies Lucca & University of Napoli Federico II since December 2022. He has received his MS degree (summa cum laude) in Computer Engineering in July 2022 from the University of Napoli Federico II. His research interests include cybersecurity, attack classification, and anomaly detection.

Federated Transfer Learning for Energy Efficiency in Smart Buildings

Enrique Marmol (Universidad de Murcia, Spain); Aurora Gonz·lez Vidal (University of Murcia, Spain); JosÈ Luis Hernandez Ramos (European Commission - Joint Research Centre (JRC), Belgium); Antonio Fernando Skarmeta Gomez (University of Murcia, Spain)

Nowadays, the generation of energy consumption models in buildings needs to address several issues. Indeed, most existing buildings lack the appropriate equipment to obtain the data required to create such models. Furthermore, the nature of energy consumption data could be correlated with additional information about people in the building that could raise privacy concerns. Based on such aspects, we propose a Federated Transfer Learning (FTL) framework to handle these buildings' data without compromising any private information in which a set of buildings are clustered according to certain characteristics. On the one hand, our works leverage the properties of Federated Learning (FL) to train an energy forecasting model using a small portion of the available buildings respecting their privacy. On the other hand, we transfer the model to the rest of the buildings by using Transfer Learning (TL). We extensively evaluate our approach, and demonstrate it improves the results of alternative scenarios where FL and TL are used separately.
Speaker Enrique Mármol Campos (University of Murcia)

Enrique Mármol Campos is a Ph.D. Student at the university of Murcia. He graduated in Mathematics in 2018. Then, in 2019, he finished the M.S. in advanced math, in the specialty of operative research and statistic, at the university of Murcia. He is currently researching on federated learning applied to cybersecurity in IoT devices.

Session Chair

Paolo Bellavista

Session GI-02

Session#2 - Performance Optimization of Next Generation Internet

10:00 AM — 11:00 AM EDT
May 20 Sat, 10:00 AM — 11:00 AM EDT

FleCom: A Flexible Congestion Control Protocol in Named Data Networking

Zhuo Li (Tianjin University, China); Hao Xun (Tianjin University, China); Yang Miao (Tianjin University, China); Weizhe Zhang (Peng Cheng Laboratory, China); Peng Luo (State Grid Hebei Electric Power Research Institute, China); Kaihua Liu (Tianjin University, China)

Congestion control has a crucial role in guaranteeing quality-of-service (QoS) in Named Data Networking (NDN). But the unique multi-source and multipath transmission characteristics make the current end-to-end multipath congestion control schemes cannot be directly employed in NDN. At present, hybrid congestion control has gradually developed into a popular mechanism. But it has to address the issues that how to respond to the actual and complex NDN scenarios, and how to dynamically adjust the size of the Interest sending window (cwnd) on the consumer side to achieve congestion control. In this paper, a flexible hybrid congestion control scheme named FleCom is proposed. Specifically, routers choose multipath forwarding or Interest shaping to mitigate congestion timely and flexibly. And consumers moderately constrain the size of cwnd to coordinate with routers. The simulation results show that FleCom consistently achieves higher total throughput than existing work. In particular, the total throughput of consumers deployed with FleCom is 109.7% higher than that of consumers deployed with PCON in the BRITE-generated topology. This value changes to 48.7% after enabling the router's in-network caching feature.
Speaker Hao Xun (Tianjin University, China)

Hao Xun received the B.Sc. degree in electronic and information engineering from Northwest A&F University in 2020, China. He is currently pursuing the M.S. degree with the School of Microelectronics, Tianjin University. His research interests in future Internet architecture.

An Efficient Topology Emulation Technology for the Space-Air-Ground Integrated Network

Xiaofeng Wang (Jiangnan University, China); Taiqian Shen (Jiangnan University, China); Yi Zhang (Jiangnan University, China); Xinyu Chen (Jiangnan University, China)

The space-air-ground integrated network (SAGIN) is a comprehensive and complex network. Network emulation plays an important role in the validation and evaluation of SAGIN's new technologies. Aiming at the characteristics of heterogeneous nodes and the large-scale and dynamic topology of SAGIN, we propose an efficient topology emulation architecture of SAGIN, named SAGINHTE-Stack. To improve usability, scalability and efficiency, we propose three technologies, the unified construction technology of the object model, static topology rapid deployment technology and topology dynamic synchronization technology. The experimental data show that unified construction technology of the object model can complete the unified model construction for various heterogeneous nodes. Static topology rapid deployment technology realizes the construction of the topology model at the millisecond level and completes the rapid deployment of the emulation topology. Topology dynamic synchronization technology realizes the parsing of topology model change behavior and the synchronization of the topology model and completes the fast synchronization of the emulation topology.
Speaker Taiqian Shen (Jiangnan University)

Taiqian Shen is a graduate student at Jiangnan University. His research interests include network emulation and Space-Air-Ground Integrated Network.

CICADA: Cloud-based Intelligent Classification and Active Defense Approach for IoT Security

Roshan L Neupane (University of Missouri-Columbia, USA); Trevor Zobrist (Southeast Missouri State University, USA); Kiran Neupane (University of Missouri, USA); Shaynoah Bedford (University of the Virgin Islands, Virgin Islands, U.S.); Shreyas Prabhudev (University of California-San Diego, USA); Trevontae Haughton (University of Missouri, USA); Jianli Pan (George Mason University, USA); Prasad Calyam (University of Missouri-Columbia, USA)

Internet of Things (IoT) devices capture and process sensitive personally identifiable information such as e.g., camera feeds/health data from enterprises and households. These devices are becoming targets of prominent attacks such as Distributed-Denial-of-Service (DDoS) and Botnets, as well as sophisticated attacks (e.g., Zero Click) that are elusive by design. There is a need for cyber deception techniques that can automate attack impact mitigation at the scale that IoT networks demand. In this paper, we present a novel cloud-based active defense approach viz., "CICADA", to detect and counter attacks that target vulnerable IoT networks. Specifically, we propose a multi-model detection engine featuring a pipeline of machine/deep learning classifiers to label inbound packet flows. In addition, we devise an edge-based defense engine that utilizes three simulated deception environments (Honeynet, Pseudocomb, and Honeyclone) with increasing pretense capabilities to deceive the attacker and lower the attack risk. Our deception environments are based on a CFO triad (cost, fidelity, observability) for designing system architectures to handle attacks with diverse detection characteristics. We evaluate the effectiveness of these architectures on an enterprise IoT network setting with a scale of thousands of devices. Our detection results show ~73% accuracy for the low observability attack (Zero Click) corresponding to the BleedingTooth exploit that allows for unauthenticated remote attacks on vulnerable devices. Furthermore, we evaluate the different deception environments based on their risk mitigation potential and associated costs. Our simulation results show that the Honeyclone is able to reduce risk by ~88% when compared to a network without any defenses.
Speaker Roshan Lal Neupane (University of Missouri-Columbia)

Roshan Lal Neupane received his MS degree in Computer Science from the University of Missouri-Columbia. He is currently pursuing Ph.D. in Computer Science at the University of Missouri-Columbia. His research interests include cybersecurity, Blockchain, cloud/edge computing, the Internet of Things, artificial intelligence, and cloud and network security. He is a student member of IEEE.

Session Chair

Kaiping Xue

Gold Sponsor

Gold Sponsor

Bronze Sponsor

Student Travel Grants

Student Travel Grants

Local Organizer

Made with in Toronto · Privacy Policy · INFOCOM 2020 · INFOCOM 2021 · INFOCOM 2022 · © 2023 Duetone Corp.