1 code implementation • 15 Jul 2018 • Dongdong She, Kexin Pei, Dave Epstein, Junfeng Yang, Baishakhi Ray, Suman Jana
However, even state-of-the-art fuzzers are not very efficient at finding hard-to-trigger software bugs.
3 code implementations • 18 May 2017 • Kexin Pei, Yinzhi Cao, Junfeng Yang, Suman Jana
First, we introduce neuron coverage for systematically measuring the parts of a DL system exercised by test inputs.
1 code implementation • 16 Dec 2020 • Kexin Pei, Zhou Xuan, Junfeng Yang, Suman Jana, Baishakhi Ray
We thus train the model to learn execution semantics from the functions' micro-traces, without any manual labeling effort.
1 code implementation • 2 Oct 2020 • Kexin Pei, Jonas Guan, David Williams-King, Junfeng Yang, Suman Jana
We present XDA, a transfer-learning-based disassembly framework that learns different contextual dependencies present in machine code and transfers this knowledge for accurate and robust disassembly.
3 code implementations • 28 Apr 2018 • Shiqi Wang, Kexin Pei, Justin Whitehouse, Junfeng Yang, Suman Jana
In this paper, we present a new direction for formally checking security properties of DNNs without using SMT solvers.
2 code implementations • NeurIPS 2018 • Shiqi Wang, Kexin Pei, Justin Whitehouse, Junfeng Yang, Suman Jana
Our approach can check different safety properties and find concrete counterexamples for networks that are 10$\times$ larger than the ones supported by existing analysis techniques.
1 code implementation • NeurIPS 2019 • Chengzhi Mao, Ziyuan Zhong, Junfeng Yang, Carl Vondrick, Baishakhi Ray
Deep networks are well-known to be fragile to adversarial attacks.
2 code implementations • 14 Dec 2022 • Chengzhi Mao, Scott Geng, Junfeng Yang, Xin Wang, Carl Vondrick
We apply this training loss to two adaption methods, model finetuning and visual prompt tuning.
1 code implementation • ICCV 2021 • Chengzhi Mao, Mia Chiquier, Hao Wang, Junfeng Yang, Carl Vondrick
We find that images contain intrinsic structure that enables the reversal of many adversarial attacks.
1 code implementation • CVPR 2023 • Chengzhi Mao, Revant Teotia, Amrutha Sundar, Sachit Menon, Junfeng Yang, Xin Wang, Carl Vondrick
We propose a ``doubly right'' object recognition benchmark, where the metric requires the model to simultaneously produce both the right labels as well as the right rationales.
1 code implementation • CVPR 2022 • Chengzhi Mao, Kevin Xia, James Wang, Hao Wang, Junfeng Yang, Elias Bareinboim, Carl Vondrick
Visual representations underlie object recognition tasks, but they often contain both robust and non-robust features.
1 code implementation • ECCV 2020 • Chengzhi Mao, Amogh Gupta, Vikram Nitin, Baishakhi Ray, Shuran Song, Junfeng Yang, Carl Vondrick
Although deep networks achieve strong accuracy on a range of computer vision benchmarks, they remain vulnerable to adversarial attacks, where imperceptible input perturbations fool the network.
1 code implementation • CVPR 2021 • Chengzhi Mao, Augustine Cha, Amogh Gupta, Hao Wang, Junfeng Yang, Carl Vondrick
We introduce a framework for learning robust visual representations that generalize to new viewpoints, backgrounds, and scene contexts.
Ranked #44 on Image Classification on ObjectNet (using extra training data)
1 code implementation • 22 Apr 2020 • Robby Costales, Chengzhi Mao, Raphael Norwitz, Bryan Kim, Junfeng Yang
We propose a live attack on deep learning systems that patches model parameters in memory to achieve predefined malicious behavior on a certain set of inputs.
1 code implementation • 16 Oct 2023 • Haozhe Chen, Junfeng Yang, Carl Vondrick, Chengzhi Mao
Large-scale pre-trained vision foundation models, such as CLIP, have become de facto backbones for various vision tasks.
1 code implementation • 23 Jan 2024 • Chengzhi Mao, Carl Vondrick, Hao Wang, Junfeng Yang
We find that large language models (LLMs) are more likely to modify human-written text than AI-generated text when tasked with rewriting.
1 code implementation • 25 Feb 2021 • Yu Jian Wu, Hongyi Wang, Yuhong Zhong, Asaf Cidon, Ryan Stutsman, Amy Tai, Junfeng Yang
The overhead of the kernel storage path accounts for half of the access latency for new NVMe storage devices.
Operating Systems Databases
1 code implementation • 7 Apr 2022 • Matthew Lawhon, Chengzhi Mao, Junfeng Yang
In this paper, we propose a novel defense that can dynamically adapt the input using the intrinsic structure from multiple self-supervised tasks.
1 code implementation • 22 Apr 2022 • Wei Hao, Aahil Awatramani, Jiayang Hu, Chengzhi Mao, Pin-Chun Chen, Eyal Cidon, Asaf Cidon, Junfeng Yang
In this paper, we introduce a new evasive attack, DIVA, that exploits these differences in edge adaptation, by adding adversarial noise to input data that maximizes the output difference between the original and adapted model.
no code implementations • 5 Dec 2017 • Kexin Pei, Linjie Zhu, Yinzhi Cao, Junfeng Yang, Carl Vondrick, Suman Jana
Finally, we show that retraining using the safety violations detected by VeriVis can reduce the average number of violations up to 60. 2%.
no code implementations • ICCV 2017 • Junfeng Yang, Xueyang Fu, Yuwen Hu, Yue Huang, Xinghao Ding, John Paisley
We incorporate domain-specific knowledge to design our PanNet architecture by focusing on the two aims of the pan-sharpening problem: spectral and spatial preservation.
no code implementations • 6 Oct 2019 • Guangyu Shen, Chengzhi Mao, Junfeng Yang, Baishakhi Ray
Due to the inherent robustness of segmentation models, traditional norm-bounded attack methods show limited effect on such type of models.
1 code implementation • 9 Nov 2020 • Jake Lee, Junfeng Yang, Zhangyang Wang
We present the results of three experiments comparing representations of millions of images with exhaustively shifted objects, examining both local invariance (within a few pixels) and global invariance (across the image frame).
no code implementations • 4 Oct 2022 • Kexin Pei, Dongdong She, Michael Wang, Scott Geng, Zhou Xuan, Yaniv David, Junfeng Yang, Suman Jana, Baishakhi Ray
Notably, NeuDep also outperforms the current state-of-the-art on these tasks.
no code implementations • 12 Nov 2022 • Victor Robila, Kexin Pei, Junfeng Yang
It is beneficial to develop an efficient machine-learning based method for addition using embedded hexadecimal digits.
1 code implementation • 12 Dec 2022 • Chengzhi Mao, Lingyu Zhang, Abhishek Joshi, Junfeng Yang, Hao Wang, Carl Vondrick
In this paper, we introduce a framework that uses the dense intrinsic constraints in natural images to robustify inference.
no code implementations • 13 Dec 2022 • Lingyu Zhang, Chengzhi Mao, Junfeng Yang, Carl Vondrick
Even under adaptive attacks where the adversary knows our defense, our algorithm is still effective.
no code implementations • 26 Dec 2022 • Pierre Tholoniat, Kelly Kostopoulou, Mosharaf Chowdhury, Asaf Cidon, Roxana Geambasu, Mathias Lécuyer, Junfeng Yang
This DP budget can be regarded as a new type of compute resource in workloads of multiple ML models training on user data.
no code implementations • 22 Mar 2023 • Yun-Yun Tsai, Ju-Chin Chao, Albert Wen, Zhaoyuan Yang, Chengzhi Mao, Tapan Shah, Junfeng Yang
Test-time defenses solve these issues but most existing test-time defenses require adapting the model weights, therefore they do not work on frozen models and complicate model memory management.
no code implementations • 12 May 2023 • Wei Hao, Zixi Wang, Lauren Hong, Lingxiao Li, Nader Karayanni, Chengzhi Mao, Junfeng Yang, Asaf Cidon
ML models are increasingly being pushed to mobile devices, for low-latency inference and offline operation.
no code implementations • 7 Aug 2023 • Kexin Pei, Weichen Li, Qirui Jin, Shuyang Liu, Scott Geng, Lorenzo Cavallaro, Junfeng Yang, Suman Jana
This paper tackles the challenge of teaching code semantics to Large Language Models (LLMs) for program analysis by incorporating code symmetries into the model architecture.
no code implementations • 29 Oct 2023 • Noah Thomas McDermott, Junfeng Yang, Chengzhi Mao
Instead, we propose to make the language models robust at test time.
no code implementations • 15 Nov 2023 • Youran Dong, Shiqian Ma, Junfeng Yang, Chao Yin
Bilevel optimization has received more and more attention recently due to its wide applications in machine learning.
no code implementations • 8 Feb 2024 • Ritambhara Singh, Abhishek Jain, Pietro Perona, Shivani Agarwal, Junfeng Yang
We have rigorously tested our method using leading-edge semantic segmentation datasets.