Research Outputs

Permanent URI for this communityhttps://hdl.handle.net/20.500.14288/2

Browse

Filters

2021 - 20244

Advanced Search

Filter by

Settings

Sponsors: search.filters.sponsors.TÜBİTAKSubject: search.filters.subject.Internet of things

Search Results

Now showing 1 - 4 of 4
  • Placeholder
    PublicationMetadata only
    Byzantines can also learn from history: fall of centered clipping in federated learning
    (IEEE-Inst Electrical Electronics Engineers Inc, 2024) Özfatura, Emre; Gündüz, Deniz; Department of Computer Engineering; Özfatura, Ahmet Kerem; Küpçü, Alptekin; Department of Computer Engineering; Koç Üniversitesi İş Bankası Enfeksiyon Hastalıkları Uygulama ve Araştırma Merkezi (EHAM) / Koç University İşbank Center for Infectious Diseases (KU-IS CID); Graduate School of Sciences and Engineering; College of Engineering;  
    The increasing popularity of the federated learning (FL) framework due to its success in a wide range of collaborative learning tasks also induces certain security concerns. Among many vulnerabilities, the risk of Byzantine attacks is of particular concern, which refers to the possibility of malicious clients participating in the learning process. Hence, a crucial objective in FL is to neutralize the potential impact of Byzantine attacks and to ensure that the final model is trustable. It has been observed that the higher the variance among the clients' models/updates, the more space there is for Byzantine attacks to be hidden. As a consequence, by utilizing momentum, and thus, reducing the variance, it is possible to weaken the strength of known Byzantine attacks. The centered clipping (CC) framework has further shown that the momentum term from the previous iteration, besides reducing the variance, can be used as a reference point to neutralize Byzantine attacks better. In this work, we first expose vulnerabilities of the CC framework, and introduce a novel attack strategy that can circumvent the defences of CC and other robust aggregators and reduce their test accuracy up to %33 on best-case scenarios in image classification tasks. Then, we propose a new robust and fast defence mechanism that is effective against the proposed and other existing Byzantine attacks.
  • Placeholder
    PublicationMetadata only
    Federated learning for pedestrian detection in vehicular networks
    (Institute of Electrical and Electronics Engineers Inc., 2023) Bennis, Mehdi; Elgabli, Anis; Gündüz, Deniz; Karaağaç, Sercan; Department of Electrical and Electronics Engineering; Kümeç, Feyzi Ege; Reyhanoğlu, Aslıhan; Kar, Emrah; Turan, Buğra; Ergen, Sinem Çöleri; Department of Electrical and Electronics Engineering; Graduate School of Sciences and Engineering; College of Engineering; Koc University Ford Otosan Automotive Technologies Laboratory (KUFOTAL)
    Vehicular connectivity is foreseen to increase road safety by enabling connected vehicle applications. On the other hand, machine learning (ML) methods are provisioned to increase road safety by supporting object detection and assisted driving. Recently, distributed ML methods, which rely on data transmission between a parameter server and vehicular edge devices, are introduced to develop intelligent transportation systems. In this paper, we investigate the feasibility of the usage of a distributed ML algorithm, federated learning (FL), to detect pedestrians by using vehicular networks. We first provide a comprehensive overview of the proposed scheme, then highlight the methodology to enable FL-based pedestrian detection from the images obtained by vehicle cameras. We further present experimental validation results for communication resource utilization, and pedestrian detection accuracy by using convolutional neural networks (CNNs) and deep neural networks (DNNs) layers in our model architecture for an FL scheme. We obtain 90% pedestrian detection accuracy with our FL scheme. © 2023 IEEE.
  • Placeholder
    PublicationMetadata only
    FLAGS simulation framework for federated learning algorithms
    (Institute of Electrical and Electronics Engineers Inc., 2023) Department of Computer Engineering; Lodhi, Ahnaf Hannan; Shamsizade, Toghrul; Al Asaad, Omar Mohammad; Akgün, Barış; Özkasap, Öznur; Department of Computer Engineering; Graduate School of Sciences and Engineering; College of Engineering
    Federated Learning (FL) provides an effective mechanism for distributed learning. However, it is expected to operate in a highly diverse setting with distinct behaviors from the participating nodes as well as dynamic network conditions. The FL performance, therefore, is subject to change due to the highly transitory nature of the overall system. An efficient simulation framework must be flexible to allow a range of participant behaviors, interactions, and environment characteristics. In this demo paper, we present the Federated Learning Algorithm Simulation (FLAGS) framework that we propose as a lightweight FL implementation and testing platform. FLAGS framework allows for a wide range of device behaviors and cooperative mechanisms, enabling rapid testing of multiple FL algorithms. © 2023 IEEE.
  • Thumbnail Image
    PublicationOpen Access
    M-stability: threshold security meets transferable utility
    (Association for Computing Machinery (ACM), 2021) Department of Computer Engineering; Biçer, Osman; Küpçü, Alptekin; Yıldız, Burcu; Faculty Member; Department of Computer Engineering; College of Engineering; Graduate School of Sciences and Engineering; N/A; 168060; N/A
    Use of game theory and mechanism design in cloud security is a well-studied topic. When applicable, it has the advantages of being efficient and simple compared to cryptography alone. Most analyses consider two-party settings, or multi-party settings where coalitions are not allowed. However, many cloud security problems that we face are in the multi-party setting and the involved parties can almost freely collaborate with each other. To formalize the study of disincentivizing coalitions from deviating strategies, a well-known definition named k-resiliency has been proposed by Abraham et al. (ACM PODC '06). Since its proposal, k-resiliency and related definitions are used extensively for mechanism design. However, in this work we observe the shortcoming of k-resiliency. That is, although this definition is secure, it is too strict to use for many cases and rule out secure mechanisms as insecure. To overcome this issue, we propose a new definition named ?.,""-repellence against the presence of a single coalition to replace k-resiliency. Our definition incorporates transferable utility in game theory as it is realistic in many distributed and multi-party computing settings. We also propose m-stability definition against the presence of multiple coalitions, which is inspired by threshold security in cryptography. We then show the advantages of our novel definitions on three mechanisms, none of which were previously analyzed against coalitions: incentivized cloud computation, forwarding data packages in ad hoc networks, and connectivity in ad hoc networks. Regarding the former, our concepts improve the proposal by Küpçü (IEEE TDSC '17), by ensuring a coalition-proof mechanism.