Joint communication and sensing (JCAS) and wireless sensing in general are gaining ever more attention in the research community. In general, wide-band sensing is required for high accuracy of the obtained distance measurements. However, wide-band sensing requires comparably expensive hardware (in terms of invest as well as energy consumption). As a solution, channel splicing has been developed for wide-band sensing using multiple narrow-band measurements. Despite many theoretical concepts, channel splicing at mmWave frequencies is still widely unexplored. In this paper, we present mmSplicer, which extends a splicing algorithm from the literature. We implemented mmSplicer in an SDR-based testbed for channel splicing at mmWave frequencies. First validation results show very high accuracy of the channel impulse response (CIR) in time domain.

The evolving unknown cyberattacks have rapidly increased the cyber threat surface. However, since only known cyberattack samples are usually available, most existing intrusion detection systems (IDSs) are only effective in detecting known cyberattacks. In this poster, we propose a cooperative differential evolution based IDS, coined CoDE-IDS, to detect unknown cyberattacks.
Specifically, we first design a dual DE algorithm for known nonself antigens (abnormal data), which considers dual evolutionary objectives and creates unknown nonself antigens characterized by similar features to known ones. Second, a hierarchical DE algorithm is designed to transfer known self antigens with hierarchical distances and create unknown nonself antigens deviating from normal patterns. Last, the newly-evolved antigens are employed to generate cyberattack detectors. Extensive experiments demonstrate that the CoDE-IDS achieves outperformed effectiveness in recognizing both unknown and known cyberattacks compared to the state-of-the-art studies.

Efficient resource scheduling in reconfigurable datacenter networks (RDCNs) is critical, but conventional schemes face challenges in responsiveness and adaptability, particularly during traffic bursts. This paper presents P4ToR, a novel distributed scheduling framework that uses P4-based programmable switches for improved traffic monitoring and dynamic optical circuit switch (OCS) allocation. P4ToR improves network responsiveness and scalability, as evident in 38.01% increase in throughput with significant reduction in flow completion time (FCT). These improvements demonstrate the potential of P4ToR as a scalable and efficient solution for RDCN resource management.

Offloading methods play a key role in optimizing computation, minimizing latency, and enhancing the overall performance of the VEC system by transfer of tasks between vehicles and edge servers or other computational resources. However, the dynamic alterations in network topology induced by the rapid mobility of vehicles, along with the emergence of topology links influenced by factors such as privacy and communication preferences, have often been overlooked. In this work, we propose a Graph reinforcement learning(GRL) method to adaptively optimize the sum of the delay of tasks in joint vehicle-to-vehicle(V2V) and vehicle-to-infrastructure(V2I). Extensive experimental results demonstrate the effectiveness and superiority of the proposed GRL-based method.

Backscatter is a promising way for ubiquitous communication. However, throughput has always been a serious bottleneck in WiFi backscatter systems due to redundant modulation. Most systems are limited to symbol-level modulation. Although Tscatter attempts to perform sample-level system, it pays a large computational cost. To solve this issue, we first show the difficulty in achieving fine-modulation lies in transmission channel influences. Based on that, we present Soscatter, a sub-symbol OFDM WiFi backscatter system achieving throughput boost by multi-step channel equalization. We are the first to show that the time-domain data samples after removing channel effects can be directly used for decoding tag data bits embedded at the sample-level. Soscatter's throughput is at least to 64x that of a symbol-level system, and 16x that of Tscatter.

Image-based teleoperation offers significant flexibility and efficiency in several applications, such as teleoperated driving; still, it highly depends on reliable communication bandwidth and high Signal-to-Noise Ratio (SNR), which is hard to guarantee in uncontrolled environments. This poster tackles the challenge of reliable communication under limited bandwidth. We propose to leverage the context and task knowledge to guide the compression so that the system can achieve good task performance rather than image fidelity. In particular, we jointly designed source-channel coding with a task performer to present an end-to-end TAsk-Guided Image Communication (TAGIC) framework, which uses Soft Introspective Variational Autoencoder (S-IntroVAE) and prioritizes the task-critical image information with limited communication bandwidth in the low SNR region. We demonstrate the effectiveness of TAGIC in a teleoperated driving scenario through the CARLA simulation platform – a widely used simulator in the autonomous driving community. Given the equivalent value of bandwidth compression ratio, TAGIC achieves a 202.6% improvement in the driving score over existing methods at low SNR.

In-band Network Telemetry (INT) is a promising technique for investigating the real-time state of the network. However, it inevitably generates considerable transmission overhead due to the operation feature of directly inserting telemetry data into the packet header. One of the widely used methods to alleviate overhead is sampling; however, it requires careful selection of sampling rates. To address this issue, we propose a novel approach that considers the frequency characteristics of each telemetry item, performs a frequency analysis on the per-packet trace, and decides a proper sampling rate for each item. We also devise an item-wise probabilistic sampling mechanism to collect items individually considering different sampling rates.

The adoption of the multipath transport protocol in the wide-area network is restricted due to the obstruction of middleboxes and its insufficient flexibility in application-level support. In this poster, we present MPP, a multipath transport paradigm that reconstructs the multipath transport protocol in user-space. Without modifying existing transport protocols (TCP, UDP, and QUIC), MPP encapsulates them into a uniform transport path that is compatible with the current network infrastructure. Multipathing functions in the multipath transport protocol are decoupled into separate modules. MPP offers APIs for developers to customize these modules with scenario-oriented multipath scheduling algorithms. The evaluation shows that our MPP prototype performs comparably with multipath transport protocols in regular scenarios and performs better in real-time video streaming with the custom scheduler we developed with MPP APIs.

Federated Learning (FL) is vulnerable to various attacks including poisoning and inference. However, the existing offensive security evaluation of FL assumes that the attackers know data distribution. In this paper, we present a novel attack where FL participants carry out inference and privacy abuse attacks against the FL by leveraging Generating Adversarial Networks (GANs). The attacker disguises the FL server as a benign FL participant. It uses GANs to generate a similar dataset as other participants and then covertly poison the data. We demonstrated the attack successfully and tested it on two datasets, the IoT dataset and MNIST. The results reveal that for FL to be successfully used in IoT applications, protection against such attacks is essential.

Mobile traffic forecasting is instrumental in efficiently managing network resources. In this poster paper, we dissect the behavior of advanced time series forecasting techniques, namely DLinear and PatchTST, when applied to the problems of predicting future mobile traffic volumes. Being black-box models hard to interpret, we ground our analysis on EXplainable Artificial Intelligence (XAI) by using AICHRONOLENS, a new tool that links legacy XAI explanations with the temporal properties of the input sequences. We find that the DLinear significantly improves the prediction accuracy over PatchTST and state-ofthe- art techniques like Long-Short Term Memory (LSTM). The analysis with AICHRONOLENS shows that, unlike PatchTST, DLinear is capable of focusing its prediction decisions on a few key samples of the input sequences, which makes it possible for DLinear to match the ground truth closely.

Elections are a cornerstone of democratic societies, and their outcome has important implications on the life of citizens and on the interior and foreign politics of a country. Understanding biases in the political orientation of the electorate plays a key role in assessing the health of the voting process and the reasons underlying the preferences of voters. Traditionally, political orientation has been studied through the lenses of the socioeconomic status of voters, e.g. their education level, type of occupation, wealth, or age. In this work, we take an original perspective and factor in mobile app usage as a different yet primary indicator of the vote decision. To this end, we explore the relationship between the 2019 European parliamentary election results in approximately 4,000 urban communes of France and the associated consumption of a wide range of mobile services. Our results show how app usage provides complementary information to the socioeconomic status and can feed a Dirichlet regression that is up to 21% more accurate in predicting the multiparty election outcome.

The satisfaction of AI computing demands relies on the robust computational foundation provided by efficient optical backbone network. Deep reinforcement learning (DRL) methods have been widely adopted recently to address elastic optical network (EON) service provisioning problems with complex spectrum constraints. However, previous frameworks utilizing path-level decision spaces face challenges as path-level features cannot be directly transferred when network topology changes, making it necessary to retrain the agent, hindering the practical deployment of the algorithm. Based on the above observations, this paper proposes a novel hop-by-hop DRL service provisioning framework for EON: Opti-DeepRoute. A node-level feature engineering is employed to construct solutions incrementally through the creation of node sequences. Utilizing the topological and spectral local receptive fields provided by the Graph Attention Network (GAT), the agent acquires local network states and partial solution information to achieve topological adaptability. Preliminary evaluations demonstrate the advantages of Opti-DeepRoute in terms of topological adaptability compared with SOTA DRL approaches.

This paper investigates a content recommendation-based edge caching method in multi-tier edge-cloud networks while considering content delivery and cache replacement decisions as well as bandwidth allocation strategies. First, we formulate the optimization problem with the goal of minimizing long-term content delivery delay. Second, we model the optimization problem as a Partially Observable Markov Decision Process, and propose a Federated Distributed Deep Deterministic Policy Gradient-based method (FD3PG) to solve the corresponding problem. In conclusion, simulation results demonstrate that the proposed FD3PG achieves not only a much lower content delivery delay but also a higher cache hit rate compared with other baselines in various scenarios.