My research focuses on trustworthy and resource-efficient machine learning over distributed and networked systems. I am particularly interested in how learning systems can provide meaningful guarantees about the use of data, remain secure during adaptation and collaborative training, and operate efficiently under communication and computation constraints.
Source-Free Machine Unlearning and Model Adaptation
As machine learning models are increasingly trained on large, distributed, and continuously changing datasets, an important question is how a trained model can remove the influence of data that should no longer be retained, such as data that should no longer be retained due to privacy or copyright concerns, or data that is outdated, incorrect, or no longer representative to the downstream task. My recent research studies machine unlearning, with a particular focus on settings where there is limited information about the original training process, such as when the original training data is no longer available. We develop source-free unlearning methods and certified approaches that aim to remove designated training information while preserving model utility.
Related publications:
Umit Yigit Basaran, Sk Miraj Ahmed, Amit K. Roy-Chowdhury, Basak Guler,
A Certified Unlearning Approach without Access to Source Data (ICML 2025).
Sk Miraj Ahmed, Umit Yigit Basaran, Dripta S. Raychaudhuri, Arindam Dutta, Rohit Kundu, Fahim Faisal Niloy, Basak Guler, Amit Roy-Chowdhury,
Towards Source-Free Machine Unlearning (CVPR 2025)
Privacy and Security of Parameter-Efficient Adaptation for Foundation Models
A second direction we study is privacy and security risks that arise when large pretrained models are adapted using private data. In particular, we investigate whether information about fine-tuning data can be recovered from shared model updates, including in parameter-efficient fine-tuning methods such as LoRA. Our recent work develops gradient-inversion attacks that expose previously unrecognized privacy vulnerabilities in distributed and federated fine-tuning.
Related publications:
Hasin Us Sami, Swapneel Sen, Basak Guler,
MineGrad: Gradient Inversion Attacks on LoRA Fine-Tuning (AISTATS 2026)
Hasin Us Sami, Swapneel Sen, Amit K. Roy-Chowdhury, Srikanth Krishnamurthy, Basak Guler,
Gradient Inversion Attacks on Parameter-Efficient Fine-Tuning (CVPR 2025)
Federated Learning under Resource Constraints and User Heterogeneity
Federated learning is a distributed learning framework that allows on-device machine learning in mobile environments. The global model is typically maintained by a central server, while the training data is kept on the user device. Rather than sending their local datasets to the server, users locally update the global model. Secure aggregation allows local updates to be aggregated at the server to update the global model, but without learning the individual updates. Unlike conventional multi-party computing mechanisms, secure aggregation should be designed to tolerate unknown user dropouts that may occur frequently in wireless networks. The global model is then pushed back to the mobile users for the next training round and/or inference. Federated learning must often operate under stringent communication, computation, and privacy constraints. Our research develops scalable federated-learning and secure aggregation methods that explicitly account for these limitations. We study how communication constraints, heterogeneous users and data, wireless resources, and security requirements affect collaborative learning, and design algorithms that remain effective under these different forms of heterogeneity. In this line of work we investigated secure aggregation for personalized and clustered federated learning, submodel training, and model sparsification.
Related publications:
Hasin Us Sami, Basak Guler Secure Gradient Aggregation with Sparsification for Resource-Limited Federated Learning, IEEE Transactions on Communications, Aug. 2024.
Hasin Us Sami, Basak Guler Secure Aggregation for Clustered Federated Learning with Passive Adversaries, IEEE Transactions on Communications, Feb. 2024.
Hasin Us Sami, Basak Guler, Secure Submodel Aggregation for Resource-Aware Federated Learning, ISIT International Symposium on Information Theory, ISIT'2024.
Hasin Us Sami, Basak Guler, Sparsity Based Secure Gradient Aggregation for Resource-Constrained Federated Learning, ISIT International Symposium on Information Theory, ISIT'2024.
Hasin Us Sami, Basak Guler, Secure Aggregation for Clustered Federated Learning, ISIT International Symposium on Information Theory, ISIT'2023.
Secure and Privacy-preserving Distributed Learning
A longstanding direction of my research is the design of distributed learning methods that provide formal privacy guarantees while remaining scalable. I develop information-theoretic approaches for secure multi-party learning, with particular emphasis on reducing communication complexity, supporting multiple rounds of training, and remaining robust to client dropouts.
Related publications:
Xingyu Lu, Umit Yigit Basaran, Basak Guler, Scalable Multi-Round Multi-Party Privacy-Preserving Neural Network Training, IEEE Transactions on Information Theory, Aug. 2024.
Xingyu Lu, Hasin Us Sami, Basak Guler, Privacy-preserving Collaborative Learning with Linear Communication Complexity, IEEE Transactions on Information Theory, Dec. 2023.
Xingyu Lu, Hasin Us Sami, Basak Guler, SCALR: Communication-Efficient Secure Multi-Party Logistic Regression, IEEE Transactions on Communications, Aug. 2023.
Umit Yigit Basaran, Xingyu Lu, Basak Guler, Beyond Secure Aggregation: Scalable Multi-Round Secure Collaborative Learning, FL-ICML Workshop on Federated Learning and Analytics in Practice: Algorithms, Systems, Applications, and Opportunities, in conjunction with International Conference on Machine Learning, ICML'2023.
Xingyu Lu, Basak Guler, Breaking the Quadratic Communication Overhead of Secure Multi-Party Neural Network Training, ISIT International Symposium on Information Theory, ISIT'2023.
Jinhyun So, Ramy E. Ali, Basak Guler, Jiantao Jiao, A. Salman Avestimehr, Securing Secure Aggregation: Mitigating Multi-Round Privacy Leakage in Federated Learning, AAAI Conference on Artificial Intelligence, AAAI 2023.
Jinhyun So, Basak Guler, A. Salman Avestimehr, Turbo-Aggregate: Breaking the Quadratic Aggregation Barrier in Secure Federated Learning, IEEE Journal on Selected Areas in Information Theory: Privacy and Security of Information Systems, Jan. 2021.
Jinhyun So, Basak Guler, A. Salman Avestimehr, CodedPrivateML: A Fast and Privacy-Preserving Framework for Distributed Machine Learning, IEEE Journal on Selected Areas in Information Theory: Privacy and Security of Information Systems, Jan. 2021
Jinhyun So, Basak Guler, A. Salman Avestimehr, A Scalable Approach for Privacy-Preserving Collaborative Machine Learning, Conference on Neural Information Processing Systems (NeurIPS 2020).
Jinhyun So, Basak Guler, A. Salman Avestimehr, BREA: Byzantine-Resilient Secure Federated Learning, IEEE Journal in Selected Areas in Communications: Machine Learning in Communications and Networks, Oct. 2020.
Learning and Secure Computation over Wireless and Edge Networks
Another direction of our work studies how distributed learning and secure computation interact with the physical constraints of wireless and edge networks. In this line of research communication is not treated merely as an abstract bottleneck, but instead we jointly consider learning, communication, computation, and resource allocation by taking into account the physical constraints of the wireless medium.
Related publications:
Hasin Us Sami, Basak Guler, Over-the-air Clustered Federated Learning, IEEE Transactions on Wireless Communications, Dec. 2023.
Yushu Yan, Kevin Chan, Ananthram Swami, Basak Guler, Resource-Aware Secure Multi-Party Edge Computation Offloading, IEEE Transactions on Communications, June. 2026.
Yushu Yan, Basak Guler, Resource-Aware Secure Computation Offloading, IEEE International Conference on Communications, ICC 2025.
Yushu Yan, Xiangyu Chang, Amit Roy-Chowdhury, Srikanth Krishnamurthy, Ananthram Swami, Basak Guler, Clustered Federated Learning for Massive MIMO Power Allocation, IEEE Military Communications Conference (MILCOM), 2024.
Yushu Yan, Xiangyu Chang, Basak Guler, Amit Roy-Chowdhury, Srikanth Krishnamurthy, Ananthram Swami, Federated Learning for Massive MIMO Power Allocation, Asilomar Conference on Signals, Systems, and Computers, 2023.
Hasin Us Sami, Basak Guler, Over-the-air Personalized Federated Learning, International Conference on Acoustics, Speech, and Signal Processing, ICASSP'21.
Network Information Theory and Joint Source-Channel Coding
My earlier research studied information-theoretic performance limits of multi-user and context-aware communication networks, including lossy distributed source-channel coding in the presence of side information. This is inspired by scenarios in which interacting parties are influenced by side information while interpreting the messages, such as external information resources or knowledge bases representing the unique backgrounds, characteristics, or biases. To do so, my research explores the fundamental limits of the amount of information that can be transferred with a fidelity criterion in a multi-user communication network when interacting parties have access to side information.
Related publications:
Basak Guler, Deniz Gündüz, Aylin Yener,Lossy Transmission of Correlated Sources over a Multiple Access Channel: Necessary Conditions and Separation Results, IEEE Transactions on Information Theory, Sep. 2018.
Basak Guler, Deniz Gündüz, Aylin Yener, On the Necessary Conditions for Transmitting Correlated Sources over a Multiple Access Channel, IEEE International Symposium on Information Theory, ISIT 2017.
Basak Guler, Deniz Gündüz, Aylin Yener,On Lossy Transmission of Correlated Sources over a Multiple Access Channel, IEEE International Symposium on Information Theory, ISIT 2016.
Basak Guler, Kaya Tutuncuoglu, Aylin Yener,Maximizing Recommender's Influence in a Social Network: An Information-Theoretic Perspective, IEEE Information Theory Workshop, ITW 2015.
Basak Guler, Aylin Yener, Ebrahim MolavianJazi, Prithwish Basu, Ananthram Swami, Carl Andersen, Interactive Function Compression with Asymmetric Priors, IEEE Data Compression Conference, DCC 2016.
Basak Guler, Aylin Yener, Compressing Semantic Information with Varying Priorities, IEEE Data Compression Conference, DCC 2014.
Semantic Communications
Emerging networks such as the Internet of Things (IoT) are designed to facilitate the interaction of humans with intelligent machines. These networks consist of actors with possibly different characteristics, goals, and interests. Such differences can in turn lead to various interpretations of the received information. I design networks that operate under such ambiguous environments, by leveraging the semantic and social features of information transmission. Unlike conventional communication networks, this necessitates taking into account the personal background and characteristics of the interacting parties. For these networks, reliable communication implies that the intended meaning of messages is preserved at reception. In effect, this new generation of networks supports interaction at a level that communicating parties can form social relationships and build trust, which may further affect how the received messages are interpreted. In contrast, communication protocols that operate in the physical layer do not take into account the difference between the meanings of transmitted and recovered messages, but rather are concerned with the engineering problem of reliably communicating sequences of bits to the receiver. These factors together motivate a new approach that molds physical and application layer metrics into one, i.e., a novel performance criterion that takes into account the meanings of the communicated messages. I designed mechanisms to achieve this, i.e., how to reliably communicate the meanings of messages through a noisy channel.
An external influential entity, who can influence how the destination perceives the received information, is considered, to model the impact of social influence on how the messages are interpreted. The exact nature of the individual, whether adversarial or helpful, is unknown to the communicating parties. An individual with such influence capability can have a significant impact on information recovery, hence transmission policies should be tailored to take into account the uncertainty in the intentions of such influential entities.
Related publications:
Basak Guler, Aylin Yener, and Ananthram Swami, The Semantic Communication Game, IEEE Transactions on Cognitive Communications and Networking, Dec. 2018.
Basak Guler, Aylin Yener, and Ananthram Swami, The Semantic Communication Game, Proceedings of the IEEE International Conference on Communications, ICC'16.
Basak Guler, Aylin Yener, Semantic Index Assignment, IEEE PERCOM Workshops 2014.
Heterogeneous Wireless Networks and Interference Management
A heterogeneous wireless network is an environment that consists of cellular base stations of various sizes, coverage areas, and operating protocols. Some of these are installed and maintained by the mobile operator, such as macrocells (cellular towers), while the others are plug-and-play devices installed by the users, such as femtocells.
This ad-hoc nature of deployment makes interference management across different tiers a challenge for mobile operators. I addressed this problem in a two-tier network that involves femtocells in addition to macrocells. Mobile user devices as well as base stations are deployed with multiple antennas, which creates spatial dimensions that can be used for interference cancellation or data rate increase. To alleviate cross-tier interference, interference received from the macrocell users can be aligned in a small dimensional subspace at multiple femtocells, while simultaneously ensuring that the performance requirements of the macrocell users are satisfied. This can enable coexistence with high data rates even at very high interference levels, when communication would be impossible otherwise.
Related publications:
Basak Guler, Aylin Yener, Uplink Interference Management in Coexisting MIMO Femtocell and Macrocell Networks: An Interference Alignment Approach, IEEE Transactions on Wireless Communications, Apr. 2014.
Basak Guler and Aylin Yener, Selective Interference Alignment for MIMO Cognitive Femtocell Networks, IEEE Journal in Selected Areas in Communications, Mar. 2014.
Basak Guler and Aylin Yener, Selective Interference Alignment for MIMO Femtocell Networks, IEEE International Conference on Communications, ICC 2013.
Basak Guler and Aylin Yener, Interference Alignment for Cooperative MIMO Femtocell Networks, Globecom 2011.