Communications & Security for Machine Learning Networks- Invited Special Session

B1L-E: Communications & Security for Machine Learning Networks- Invited Special Session

Session Type: Lecture
Session Code: B1L-E
Location: Room 5
Date & Time: Thursday March 23, 2023 (09:00 - 10:00)
Chair: Arlene Cole-Rhodes
Track: 12
Paper No. Paper NameAuthorsAbstract
3048An Overview of Quantum-Safe Approaches: Quantum Key Distribution and Post-Quantum CryptographyGuobin Xu, Jianzhou Mao, Eric Sakk, Shuangbao WangCommon cryptographic algorithms may no longer be considered secure under future quantum computers, which will cause a serious threat to network security. Hence, investigating quantum-safe cryptography and evaluating the safety of traditional cryptographic algorithms are essential and have become urgent demands. In this paper, we study the quantum-safe cryptography approaches and conduct a survey of the various quantum key distribution protocols, simulation tools, and commercial applications. In addition, we provide a comparison of the first four post-quantum cryptographic algorithms recently announced by the National Institute of Standards and Technology. The challenges of quantum-safe approaches will be discussed, which aim to find future research directions in quantum cryptography.
3114Towards Equalization of Mixed Multi-User OFDM Signals Over a Doubly-Dispersive ChannelKirsten Toland, Peter Taiwo, Arlene Cole-RhodesWe explore the recovery of mixed signals, which have been transmitted by multiple users over a doubly-dispersive channel using OFDM modulation and received by a multi-antenna system. We consider multiple users transmitting QPSK signal blocks over a shared OFDM frequency band and we consider two cases, with respect to the location of signal data sources. The first case will involve collocated users transmitting blocks of QPSK signals to be received by the multi-antenna system, and the second case will involve non-collocated mobile users transmitting QPSK signal blocks to the multi-antenna system. On each OFDM subcarrier, transmitted symbols from multiple users are mixed thereby providing a low probability of interception (LPI) scenario. A modified OFDM receiver architecture will be proposed and used to separate and recover these user signals at the base station. We apply a Least Square equalizer to the signal data blocks at the receiver. These received signal blocks have also been corrupted by additive white Gaussian noise (AWGN) and inter-symbol interference (ISI) after transmission through a Rayleigh fading Doppler channel. We estimate the channel state information using pilots and measure the performance of the proposed multi-user OFDM receiver.
3135Trust Evaluation in Federated LearningSyeda Sanjidah{2}, Md Tanvir Arafin{1}Establishing and ensuring trust has become an essential aspect of machine learning. In federated learning, several clients train a central initial machine learning model with their local data, while the central server creates a global model by aggregating the local models and sending that to the clients. A number of clients could be malicious, or an attacker could manipulate the client\'s data. The data manipulation could compromise a global model and violates trustworthiness. Establishing trust in the clients is crucial for the central server, as any client on the loop could be malicious. Hence, in this work, we use a modified nearest neighbor classifier-based approach to determine the trust score of local models and the aggregated model in the FL setting to increase the trustworthiness of participating clients in a dynamic learning environment.
3163AI/ML Systems Engineering Workbench FrameworkKofi Nyarko, Peter Taiwo, Chukwuemeka Duru, Emmanuel Masa-IbiThis paper presents the framework of a cloud-based Artificial Intelligence (AI) and Machine Learning (ML) workbench that provides services utilization and performance benchmarking. The framework promotes convenience by enabling a centralized platform for software developers and data scientists to perform federated search across various dataset repositories, choose problem domains, like Natural Language Processing, Speech and Computer Vision, and build/validate models. The benchmarking functionality of this framework helps users evaluate and compare performances of various solutions from multiple cloud service providers. The workbench framework consists of two primary layers. The Services layer which is rendered as an AI as a service (AIaaS) model, providing interfaces that connect users to vision, speech and natural language processing (NLP) services from various AI service providers. The Platform layer is an ML as a Service (MLaaS) model providing access to ML model training, tuning, inference and transfer learning tasks that are fulfillable on multiple cloud ML platforms with preset cloud-based compute instances. Benchmarking is provided on the workbench by comparing accuracy metrics on prediction and detection counts, F1 scores and ML training instances setup and completion time. By utilizing these performance benchmarking features, this framework can assist AI and ML practitioners in making informed judgments when selecting a cloud provider for specific activities. Additionally, it will increase the effectiveness and efficiency of data science training for both teachers and students.