AIConfPaper
← Search

Yigitcan Kaya

8 accepted papers

2025

PoisonedParrot: Subtle Data Poisoning Attacks to Elicit Copyright-Infringing Content from Large Language Models

Michael-Andrei Panaitescu-Liess, Pankayaraj Pathmanathan, Yigitcan Kaya, Zora Che, Bang An, Sicheng Zhu +2
NAACL 2025long

As the capabilities of large language models (LLMs) continue to expand, their usage has become increasingly prevalent. However, as reflected in numerous ongoing lawsuits regarding LLM-generated content, addressing copyright infringement remains a significant challenge. In this paper, we introduce Po…

Cited by 1Source ↗PDF ↗Scholar ↗
2024

DRSM: De-Randomized Smoothing on Malware Classifier Providing Certified Robustness

Shoumik Saha, Wenxiao Wang, Yigitcan Kaya, Soheil Feizi, Tudor Dumitras
ICLR 2024poster

Machine Learning (ML) models have been utilized for malware detection for over two decades. Consequently, this ignited an ongoing arms race between malware authors and antivirus systems, compelling researchers to propose defenses for malware-detection models against evasion attacks. However, most if…

Cited by 3Source ↗PDF ↗code ↗Scholar ↗
2024

Like Oil and Water: Group Robustness Methods and Poisoning Defenses May Be at Odds

Michael-Andrei Panaitescu-Liess, Yigitcan Kaya, Sicheng Zhu, Furong Huang, Tudor Dumitras
ICLR 2024poster

Group robustness has become a major concern in machine learning (ML) as conventional training paradigms were found to produce high error on minority groups. Without explicit group annotations, proposed solutions rely on heuristics that aim to identify and then amplify the minority samples during tra…

Cited by 0Source ↗PDF ↗Scholar ↗
2022

Generating Distributional Adversarial Examples to Evade Statistical Detectors

Yigitcan Kaya, Muhammad Bilal Zafar, Sergul Aydore, Nathalie Rauschmayr, Krishnaram Kenthapadi
ICML 2022spotlight

Deep neural networks (DNNs) are known to be highly vulnerable to adversarial examples (AEs) that include malicious perturbations. Assumptions about the statistical differences between natural and adversarial inputs are commonplace in many detection techniques. As a best practice, AE detectors are ev…

Cited by 7Source ↗PDF ↗Scholar ↗
2021

A Panda? No, It's a Sloth: Slowdown Attacks on Adaptive Multi-Exit Neural Network Inference

Sanghyun Hong, Yigitcan Kaya, Ionuț-Vlad Modoranu, Tudor Dumitras
ICLR 2021spotlight

Recent increases in the computational demands of deep neural networks (DNNs), combined with the observation that most input samples require only simple models, have sparked interest in input-adaptive multi-exit architectures, such as MSDNets or Shallow-Deep Networks. These architectures enable faste…

Cited by 82Source ↗PDF ↗code ↗Scholar ↗
2021

Qu-ANTI-zation: Exploiting Quantization Artifacts for Achieving Adversarial Outcomes

Sanghyun Hong, Michael-Andrei Panaitescu-Liess, Yigitcan Kaya, Tudor Dumitras
NeurIPS 2021poster

Quantization is a popular technique that transforms the parameter representation of a neural network from floating-point numbers into lower-precision ones (e.g., 8-bit integers). It reduces the memory footprint and the computational cost at inference, facilitating the deployment of resource-hungry m…

Cited by 21Source ↗PDF ↗code ↗Scholar ↗
2021

When Does Data Augmentation Help With Membership Inference Attacks?

Yigitcan Kaya, Tudor Dumitras
ICML 2021spotlight

Deep learning models often raise privacy concerns as they leak information about their training data. This leakage enables membership inference attacks (MIA) that can identify whether a data point was in a model’s training set. Research shows that some ’data augmentation’ mechanisms may reduce the r…

Cited by 87Source ↗PDF ↗code ↗Scholar ↗
2019

Shallow-Deep Networks: Understanding and Mitigating Network Overthinking

Yigitcan Kaya, Sanghyun Hong, Tudor Dumitras
ICML 2019oral

We characterize a prevalent weakness of deep neural networks (DNNs), ’overthinking’, which occurs when a DNN can reach correct predictions before its final layer. Overthinking is computationally wasteful, and it can also be destructive when, by the final layer, a correct prediction changes into a mi…

Cited by 368Source ↗PDF ↗code ↗Scholar ↗