Low-Rank Adaptation (LoRA) is a popular technique for efficient fine-tuning of foundation models.However, applying LoRA in federated learning environments, where data is distributed across multiple clients, presents unique challenges.Existing methods rely on traditional federated averaging of LoRA adapters, resulting in inexact updates.To address this, we propose Federated Exact LoRA, or FedEx-LoRA, which adds a residual error term to the pre-trained frozen weight matrix.Our approach achieves exact updates with minimal computational and communication overhead, preserving LoRA's efficiency.We evaluate the method on various models across arithmetic reasoning, commonsense reasoning, natural language understanding and natural language generation tasks, showing consistent performance gains over state-of-the-art methods across multiple settings.Through extensive analysis, we quantify that the deviations in updates from the ideal solution are significant, highlighting the need for exact aggregation.Our method's simplicity, efficiency, and broad applicability position it as a promising solution for accurate and effective federated fine-tuning of foundation models.Our code is available

The acquisition of training data is crucial for machine learning applications. Data markets can increase the supply of data, particularly in data-scarce domains such as healthcare, by incentivizing potential data providers to join the market. A major challenge for a data buyer in such a market is choosing the most valuable data points from a data seller. Unlike prior work in data valuation, which assumes centralized data access, we propose a federated approach to the data acquisition problem that is inspired by linear experimental design. Our proposed data acquisition method achieves lower prediction error without requiring labeled validation data and can be optimized in a fast and federated procedure. The key insight of our work is that a method that directly estimates the benefit of acquiring data for test set prediction is particularly compatible with a decentralized market setting.

In recent years, split learning (SL) has emerged as a promising distributed learning framework that can utilize big data in parallel without privacy leakage while reducing client-side computing resources. In the initial implementation of SL, however, the server serves multiple clients sequentially incurring high latency. Parallel implementation of SL can alleviate this latency problem, but existing Parallel SL algorithms compromise scalability due to its fundamental structural problem. To this end, our previous works have proposed two scalable Parallel SL algorithms, dubbed SGLR and LocFedMix-SL, by solving the aforementioned fundamental problem of the Parallel SL structure. In this article, we propose a novel Parallel SL framework, coined Mix2SFL, that can ameliorate both accuracy and communication-efficiency while still ensuring scalability. Mix2SFL first supplies more samples to the server through a manifold mixup between the smashed data uploaded to the server as in SmashMix of LocFedMix-SL, and then averages the split-layer gradient as in GradMix of SGLR, followed by local model aggregation as in SFL. Numerical evaluation corroborates that Mix2SFL achieves improved performance in both accuracy and latency compared to the state-of-the-art SL algorithm with scalability guarantees. Moreover, its convergence speed as well as privacy guarantee are validated through the experimental results.

Cloud-based machine learning inference is an emerging paradigm where users query by sending their data through a service provider who runs an ML model on that data and returns back the answer. Due to increased concerns over data privacy, recent works have proposed Collaborative Inference (CI) to learn a privacy-preserving encoding of sensitive user data before it is shared with an untrusted service provider. Existing works so far evaluate the privacy of these encodings through empirical reconstruction attacks. In this work, we develop a new framework that provides formal privacy guarantees for an arbitrarily trained neural network by linking its local Lipschitz constant with its local sensitivity. To guarantee privacy using local sensitivity, we extend the Propose-Test-Release (PTR) framework to make it tractable for neural network queries. We verify the efficacy of our framework experimentally on real-world datasets and elucidate the role of Adversarial Representation Learning (ARL) in improving the privacy-utility trade-off.

Differential Privacy offers strong guarantees such as immutable privacy under any post-processing. In this work, we propose a differentially private mechanism called PrivateMail for performing supervised manifold learning. We then apply it to the use case of private image retrieval to obtain nearest matches to a client’s target image from a server’s database. PrivateMail releases the target image as part of a differentially private manifold embedding. We give bounds on the global sensitivity of the manifold learning map in order to obfuscate and release embeddings with differential privacy inducing noise. We show that PrivateMail obtains a substantially better performance in terms of the privacy-utility trade off in comparison to several baselines on various datasets. We share code for applying PrivateMail at http://tiny.cc/PrivateMail.

Split learning (SL) is a distributed deep-learning approach that enables individual data owners to train a shared model over their joint data without exchanging it with one another. SL has been the subject of much research in recent years, leading to the development of several versions for facilitating distributed learning. However, the majority of this work mainly focuses on optimizing the training process while largely ignoring the design and implementation of practical tool support. To fill this gap, we present our automated software framework for training deep neural networks from decentralized data based on our extended version of SL, termed Blind Learning. Specifically, we shed light on the underlying optimization algorithm, explain the design and implementation details of our framework, and present our preliminary evaluation results. We demonstrate that Blind Learning is 65% more computationally efficient than SL and can produce better performing models. Moreover, we show that running the same job in our framework is at least 4.5× faster than PySyft. Our goal is to spur the development of proper tool support for distributed deep learning.

Split learning (SL) is a promising distributed learning framework that enables to utilize the huge data and parallel computing resources of mobile devices. SL is built upon a model-split architecture, wherein a server stores an upper model segment that is shared by different mobile clients storing its lower model segments. Without exchanging raw data, SL achieves high accuracy and fast convergence by only uploading smashed data from clients and downloading global gradients from the server. Nonetheless, the original implementation of SL sequentially serves multiple clients, incurring high latency with many clients. A parallel implementation of SL has great potential in reducing latency, yet existing parallel SL algorithms resort to compromising scalability and/or convergence speed. Motivated by this, the goal of this article is to develop a scalable parallel SL algorithm with fast convergence and low latency. As a first step, we identify that the fundamental bottleneck of existing parallel SL comes from the model-split and parallel computing architectures, under which the server-client model updates are often imbalanced, and the client models are prone to detach from the server’s model. To fix this problem, by carefully integrating local parallelism, federated learning, and mixup augmentation techniques, we propose a novel parallel SL framework, coined LocFedMix-SL. Simulation results corroborate that LocFedMix-SL achieves improved scalability, convergence speed, and latency, compared to sequential SL as well as the state-of-the-art parallel SL algorithms such as SplitFed and LocSplitFed.

Recent deep learning models have shown remarkable performance in image classification. While these deep learning systems are getting closer to practical deployment, the common assumption made about data is that it does not carry any sensitive information. This assumption may not hold for many practical cases, especially in the domain where an individual’s personal information is involved, like healthcare and facial recognition systems. We posit that selectively removing features in this latent space can protect the sensitive information and provide better privacy-utility trade-off. Consequently, we propose DISCO which learns a dynamic and data driven pruning filter to selectively obfuscate sensitive information in the feature space. We propose diverse attack schemes for sensitive inputs & attributes and demonstrate the effectiveness of DISCO against state-of-the-art methods through quantitative and qualitative evaluation. Finally, we also release an evaluation benchmark dataset of 1 million sensitive representations to encourage rigorous exploration of novel attack and defense schemes at https://github.com/splitlearning/InferenceBenchmark.

For models trained on-premise but deployed in a distributed fashion across multiple entities, we demonstrate that minimizing distance correlation between sensitive data such as faces and intermediary representations enables prediction while preventing reconstruction attacks. Leakage (measured using distance correlation between input and intermediate representations) is the risk associated with the reconstruction of raw face data from intermediary representations that are communicated in a distributed setting. We demonstrate on face datasets that our method is resilient to reconstruction attacks during distributed inference while maintaining information required to sustain good classification accuracy. We share modular code for performing NoPeek-Infer at http://tiny.cc/nopeek along with corresponding trained models for benchmarking attack techniques.

Wireless channels can be inherently privacy preserving by distorting the received signals due to channel noise, and superpositioning multiple signals over-the-air. By harnessing these natural distortions and superpositions by wireless channels, we propose a novel privacy-preserving machine learning (ML) framework at the network edge, coined over-the-air mixup ML (AirMixML). In AirMixML, multiple workers transmit analog-modulated signals of their private data samples to an edge server who trains an ML model using the received noisy-and-superpositioned samples. AirMixML coincides with model training using mixup data augmentation achieving comparable accuracy to that with raw data samples. From a privacy perspective, AirMixML is a differentially private (DP) mechanism limiting the disclosure of each worker's private sample information at the server, while the worker's transmit power determines the privacy disclosure level. To this end, we develop a fractional channel-inversion power control (PC) method, a-Dirichlet mixup PC (DirMix(a)-PC), wherein for a given global power scaling factor after channel inversion, each worker's local power contribution to the superpositioned signal is controlled by the Dirichlet dispersion ratio a. Mathematically, we derive a closed-form expression clarifying the relationship between the local and global PC factors to guarantee a target DP level. By simulations, we provide DirMix(α)-PC design guidelines to improve accuracy, privacy, and energy-efficiency. Finally, AirMixML with DirMix(a)-PC is shown to achieve reasonable accuracy compared to a privacy-violating baseline with neither superposition nor PC.

In this work we present A-Wristocracy, a novel framework for recognizing very fine-grained and complex inhome activities of human users (particularly elderly people) with wrist-worn device sensing. Our designed A-Wristocracy system improves upon the state-of-the-art works on in-home activity recognition using wearables. These works are mostly able to detect coarse-grained ADLs (Activities of Daily Living) but not large number of fine-grained and complex IADLs (Instrumental Activities of Daily Living). These are also not able to distinguish similar activities but with different context (such as sit on floor vs. sit on bed vs. sit on sofa). Our solution helps accurate detection of in-home ADLs/ IADLs and contextual activities, which are all critically important for remote elderly care in tracking their physical and cognitive capabilities. A-Wristocracy makes it feasible to classify large number of fine-grained and complex activities, through Deep Learning based data analytics and exploiting multi-modal sensing on wrist-worn device. It exploits minimal functionality from very light additional infrastructure (through only few Bluetooth beacons), for coarse level location context. A-Wristocracy preserves direct user privacy by excluding camera/ video imaging on wearable or infrastructure. The classification procedure consists of practical feature set extraction from multi-modal wearable sensor suites, followed by Deep Learning based supervised fine-level classification algorithm. We have collected exhaustive home-based ADLs and IADLs data from multiple users. Our designed classifier is validated to be able to recognize very fine-grained complex 22 daily activities (much larger number than 6-12 activities detected by state-of-the-art works using wearable and no camera/ video) with high average test accuracies of 90% or more for two users in two different home environments.