arXiv subjects

Burak Hasircioglu

Publications and source records attributed to Burak Hasircioglu.

12 recordsLinked to original sources

arXiv2025

Poisoning Attacks on LLMs Require a Near-constant Number of Poison Samples

Poisoning attacks can compromise the safety of large language models (LLMs) by injecting malicious documents into their training data. Existing work has studied pretraining poisoning assuming adversaries control a percentage of the training corpus. However, for large models, even small percentages translate to impractically large amounts of data. This work demonstrates for the first time that poisoning attacks instead require a near-constant number of documents regardless of dataset size. We conduct the largest pretraining poisoning experiments to date, pretraining models from 600M to 13B parameters on chinchilla-optimal datasets (6B to 260B tokens). We find that 250 poisoned documents similarly compromise models across all model and dataset sizes, despite the largest models training on more than 20 times more clean data. We also run smaller-scale experiments to ablate factors that could influence attack success, including broader ratios of poisoned to clean data and non-random distributions of poisoned samples. Finally, we demonstrate the same dynamics for poisoning during fine-tuning. Altogether, our results suggest that injecting backdoors through data poisoning may be easier for large models than previously believed as the number of poisons required does not scale up with model size, highlighting the need for more research on defences to mitigate this risk in future models.

cs.LG↗

arXiv2025

From Promise to Peril: Rethinking Cybersecurity Red and Blue Teaming in the Age of LLMs

Large Language Models (LLMs) are set to reshape cybersecurity by augmenting red and blue team operations. Red teams can exploit LLMs to plan attacks, craft phishing content, simulate adversaries, and generate exploit code. Conversely, blue teams may deploy them for threat intelligence synthesis, root cause analysis, and streamlined documentation. This dual capability introduces both transformative potential and serious risks. This position paper maps LLM applications across cybersecurity frameworks such as MITRE ATT&CK and the NIST Cybersecurity Framework (CSF), offering a structured view of their current utility and limitations. While LLMs demonstrate fluency and versatility across various tasks, they remain fragile in high-stakes, context-heavy environments. Key limitations include hallucinations, limited context retention, poor reasoning, and sensitivity to prompts, which undermine their reliability in operational settings. Moreover, real-world integration raises concerns around dual-use risks, adversarial misuse, and diminished human oversight. Malicious actors could exploit LLMs to automate reconnaissance, obscure attack vectors, and lower the technical threshold for executing sophisticated attacks. To ensure safer adoption, we recommend maintaining human-in-the-loop oversight, enhancing model explainability, integrating privacy-preserving mechanisms, and building systems robust to adversarial exploitation. As organizations increasingly adopt AI driven cybersecurity, a nuanced understanding of LLMs' risks and operational impacts is critical to securing their defensive value while mitigating unintended consequences.

cs.CR↗

arXiv2025

One Pic is All it Takes: Poisoning Visual Document Retrieval Augmented Generation with a Single Image

Retrieval-augmented generation (RAG) is instrumental for inhibiting hallucinations in large language models (LLMs) through the use of a factual knowledge base (KB). Although PDF documents are prominent sources of knowledge, text-based RAG pipelines are ineffective at capturing their rich multi-modal information. In contrast, visual document RAG (VD-RAG) uses screenshots of document pages as the KB, which has been shown to achieve state-of-the-art results. However, by introducing the image modality, VD-RAG introduces new attack vectors for adversaries to disrupt the system by injecting malicious documents into the KB. In this paper, we demonstrate the vulnerability of VD-RAG to poisoning attacks targeting both retrieval and generation. We define two attack objectives and demonstrate that both can be realized by injecting only a single adversarial image into the KB. Firstly, we introduce a targeted attack against one or a group of queries with the goal of spreading targeted disinformation. Secondly, we present a universal attack that, for any potential user query, influences the response to cause a denial-of-service in the VD-RAG system. We investigate the two attack objectives under both white-box and black-box assumptions, employing a multi-objective gradient-based optimization approach as well as prompting state-of-the-art generative models. Using two visual document datasets, a diverse set of state-of-the-art retrievers (embedding models) and generators (vision language models), we show VD-RAG is vulnerable to poisoning attacks in both the targeted and universal settings, yet demonstrating robustness to black-box attacks in the universal setting.

cs.CL↗

arXiv2024

Private Collaborative Edge Inference via Over-the-Air Computation

We consider collaborative inference at the wireless edge, where each client's model is trained independently on its local dataset. Clients are queried in parallel to make an accurate decision collaboratively. In addition to maximizing the inference accuracy, we also want to ensure the privacy of local models. To this end, we leverage the superposition property of the multiple access channel to implement bandwidth-efficient multi-user inference methods. We propose different methods for ensemble and multi-view classification that exploit over-the-air computation (OAC). We show that these schemes perform better than their orthogonal counterparts with statistically significant differences while using fewer resources and providing privacy guarantees. We also provide experimental results verifying the benefits of the proposed OAC approach to multi-user inference, and perform an ablation study to demonstrate the effectiveness of our design choices. We share the source code of the framework publicly on Github to facilitate further research and reproducibility.

cs.LG↗

arXiv2023

Communication Efficient Private Federated Learning Using Dithering

The task of preserving privacy while ensuring efficient communication is a fundamental challenge in federated learning. In this work, we tackle this challenge in the trusted aggregator model, and propose a solution that achieves both objectives simultaneously. We show that employing a quantization scheme based on subtractive dithering at the clients can effectively replicate the normal noise addition process at the aggregator. This implies that we can guarantee the same level of differential privacy against other clients while substantially reducing the amount of communication required, as opposed to transmitting full precision gradients and using central noise addition. We also experimentally demonstrate that the accuracy of our proposed approach matches that of the full precision gradient method.

cs.LG↗

arXiv2022

Over-the-Air Ensemble Inference with Model Privacy

We consider distributed inference at the wireless edge, where multiple clients with an ensemble of models, each trained independently on a local dataset, are queried in parallel to make an accurate decision on a new sample. In addition to maximizing inference accuracy, we also want to maximize the privacy of local models. We exploit the superposition property of the air to implement bandwidth-efficient ensemble inference methods. We introduce different over-the-air ensemble methods and show that these schemes perform significantly better than their orthogonal counterparts, while using less resources and providing privacy guarantees. We also provide experimental results verifying the benefits of the proposed over-the-air inference approach, whose source code is shared publicly on Github.

cs.LG↗

arXiv2022

Privacy Amplification via Random Participation in Federated Learning

Running a randomized algorithm on a subsampled dataset instead of the entire dataset amplifies differential privacy guarantees. In this work, in a federated setting, we consider random participation of the clients in addition to subsampling their local datasets. Since such random participation of the clients creates correlation among the samples of the same client in their subsampling, we analyze the corresponding privacy amplification via non-uniform subsampling. We show that when the size of the local datasets is small, the privacy guarantees via random participation is close to those of the centralized setting, in which the entire dataset is located in a single host and subsampled. On the other hand, when the local datasets are large, observing the output of the algorithm may disclose the identities of the sampled clients with high confidence. Our analysis reveals that, even in this case, privacy guarantees via random participation outperform those via only local subsampling.

cs.LG↗

arXiv2022

Bivariate Polynomial Codes for Secure Distributed Matrix Multiplication

We consider the problem of secure distributed matrix multiplication (SDMM). Coded computation has been shown to be an effective solution in distributed matrix multiplication, both providing privacy against workers and boosting the computation speed by efficiently mitigating stragglers. In this work, we present a non-direct secure extension of the recently introduced bivariate polynomial codes. Bivariate polynomial codes have been shown to be able to further speed up distributed matrix multiplication by exploiting the partial work done by the stragglers rather than completely ignoring them while reducing the upload communication cost and/or the workers' storage's capacity needs. We show that, especially for upload communication or storage constrained settings, the proposed approach reduces the average computation time of SDMM compared to its competitors in the literature.

cs.IT↗

arXiv2021

Bivariate Polynomial Coding for Efficient Distributed Matrix Multiplication

Coded computing is an effective technique to mitigate "stragglers" in large-scale and distributed matrix multiplication. In particular, univariate polynomial codes have been shown to be effective in straggler mitigation by making the computation time depend only on the fastest workers. However, these schemes completely ignore the work done by the straggling workers resulting in a waste of computational resources. To reduce the amount of work left unfinished at workers, one can further decompose the matrix multiplication task into smaller sub-tasks, and assign multiple sub-tasks to each worker, possibly heterogeneously, to better fit their particular storage and computation capacities. In this work, we propose a novel family of bivariate polynomial codes to efficiently exploit the work carried out by straggling workers. We show that bivariate polynomial codes bring significant advantages in terms of upload communication costs and storage efficiency, measured in terms of the number of sub-tasks that can be computed per worker. We propose two bivariate polynomial coding schemes. The first one exploits the fact that bivariate interpolation is always possible on a rectangular grid of evaluation points. We obtain such points at the cost of adding some redundant computations. For the second scheme, we relax the decoding constraints and require decodability for almost all choices of the evaluation points. We present interpolation sets satisfying such decodability conditions for certain storage configurations of workers. Our numerical results show that bivariate polynomial coding considerably reduces the average computation time of distributed matrix multiplication. We believe this work opens up a new class of previously unexplored coding schemes for efficient coded distributed computation.

cs.IT↗

arXiv2021

Speeding Up Private Distributed Matrix Multiplication via Bivariate Polynomial Codes

We consider the problem of private distributed matrix multiplication under limited resources. Coded computation has been shown to be an effective solution in distributed matrix multiplication, both providing privacy against the workers and boosting the computation speed by efficiently mitigating stragglers. In this work, we propose the use of recently-introduced bivariate polynomial codes to further speed up private distributed matrix multiplication by exploiting the partial work done by the stragglers rather than completely ignoring them. We show that the proposed approach reduces the average computation time of private distributed matrix multiplication compared to its competitors in the literature while improving the upload communication cost and the workers' storage efficiency.

cs.IT↗

arXiv2021

Private Wireless Federated Learning with Anonymous Over-the-Air Computation

In conventional federated learning (FL), differential privacy (DP) guarantees can be obtained by injecting additional noise to local model updates before transmitting to the parameter server (PS). In the wireless FL scenario, we show that the privacy of the system can be boosted by exploiting over-the-air computation (OAC) and anonymizing the transmitting devices. In OAC, devices transmit their model updates simultaneously and in an uncoded fashion, resulting in a much more efficient use of the available spectrum. We further exploit OAC to provide anonymity for the transmitting devices. The proposed approach improves the performance of private wireless FL by reducing the amount of noise that must be injected.

cs.CR↗

arXiv2021

Dopamine: Differentially Private Federated Learning on Medical Data

While rich medical datasets are hosted in hospitals distributed across the world, concerns on patients' privacy is a barrier against using such data to train deep neural networks (DNNs) for medical diagnostics. We propose Dopamine, a system to train DNNs on distributed datasets, which employs federated learning (FL) with differentially-private stochastic gradient descent (DPSGD), and, in combination with secure aggregation, can establish a better trade-off between differential privacy (DP) guarantee and DNN's accuracy than other approaches. Results on a diabetic retinopathy~(DR) task show that Dopamine provides a DP guarantee close to the centralized training counterpart, while achieving a better classification accuracy than FL with parallel DP where DPSGD is applied without coordination. Code is available at https://github.com/ipc-lab/private-ml-for-health.

cs.LG↗