Connectome-based machine learning models are vulnerable to subtle data manipulations

被引:5
|
作者
Rosenblatt, Matthew [1 ]
Rodriguez, Raimundo X. [2 ]
Westwater, Margaret L. [3 ]
Dai, Wei [4 ]
Horien, Corey [2 ]
Greene, Abigail S. [2 ]
Constable, R. Todd [1 ,2 ,3 ,5 ]
Noble, Stephanie [3 ]
Scheinost, Dustin [1 ,2 ,3 ,6 ,7 ,8 ]
机构
[1] Yale Sch Engn & Appl Sci, Dept Biomed Engn, New Haven, CT 06510 USA
[2] Yale Sch Med, Interdept Neurosci Program, New Haven, CT 06510 USA
[3] Yale Sch Med, Dept Radiol & Biomed Imaging, New Haven, CT 06510 USA
[4] Yale Sch Publ Hlth, Dept Biostat, New Haven, CT 06510 USA
[5] Yale Sch Med, Dept Neurosurg, New Haven, CT 06510 USA
[6] Yale Univ, Dept Stat & Data Sci, New Haven, CT 06510 USA
[7] Yale Sch Med, Child Study Ctr, New Haven, CT 06510 USA
[8] Yale Univ, Wu Tsai Inst, New Haven, CT 06510 USA
来源
PATTERNS | 2023年 / 4卷 / 07期
基金
中国国家自然科学基金; 美国国家卫生研究院;
关键词
HUMAN BRAIN;
D O I
10.1016/j.patter.2023.100756
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Neuroimaging-based predictive models continue to improve in performance, yet a widely overlooked aspect of these models is "trustworthiness,"or robustness to data manipulations. High trustworthiness is imperative for researchers to have confidence in their findings and interpretations. In this work, we used functional connectomes to explore how minor data manipulations influence machine learning predic-tions. These manipulations included a method to falsely enhance prediction performance and adversarial noise attacks designed to degrade performance. Although these data manipulations drastically changed model performance, the original and manipulated data were extremely similar (r = 0.99) and did not affect other downstream analysis. Essentially, connectome data could be inconspicuously modified to achieve any desired prediction performance. Overall, our enhancement attacks and evaluation of existing adversarial noise attacks in connectome-based models highlight the need for counter-measures that improve the trustworthiness to preserve the integrity of academic research and any potential translational applications.
引用
收藏
页数:14
相关论文
共 50 条
  • [41] Stock price crash prediction based on multimodal data machine learning models
    Sheng, Yankai
    Qu, Yuanyu
    Ma, Ding
    FINANCE RESEARCH LETTERS, 2024, 62
  • [42] A Machine Learning Approach to Detection of Trade-Based Manipulations in Borsa Istanbul
    Nurullah Celal Uslu
    Fuat Akal
    Computational Economics, 2022, 60 : 25 - 45
  • [43] A Machine Learning Approach to Detection of Trade-Based Manipulations in Borsa Istanbul
    Uslu, Nurullah Celal
    Akal, Fuat
    COMPUTATIONAL ECONOMICS, 2022, 60 (01) : 25 - 45
  • [44] A Comparative Study of Shallow Machine Learning Models and Deep Learning Models for Landslide Susceptibility Assessment Based on Imbalanced Data
    Xu, Shiluo
    Song, Yingxu
    Hao, Xiulan
    FORESTS, 2022, 13 (11):
  • [45] Analytics of Epidemiological Data using Machine Learning Models
    Barapatre, Harshita
    Jangir, Jatin
    Bajpai, Sudhanshu
    Chawla, Bhavesh
    Keswani, Gunjan
    INTERNATIONAL JOURNAL OF NEXT-GENERATION COMPUTING, 2023, 14 (01): : 255 - 262
  • [46] Predicting Personality with Twitter Data and Machine Learning Models
    Ergu, Izel
    Isik, Zerrin
    Yankayis, Ismail
    2019 INNOVATIONS IN INTELLIGENT SYSTEMS AND APPLICATIONS CONFERENCE (ASYU), 2019, : 386 - 390
  • [47] Approximate Data Deletion from Machine Learning Models
    Izzo, Zachary
    Smart, Mary Anne
    Chaudhuri, Kamalika
    Zou, James
    24TH INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND STATISTICS (AISTATS), 2021, 130
  • [48] Exploring Data Distributions in Machine Learning Models with SOMs
    Konig, Caroline
    Vellido, Alfredo
    ADVANCES IN SELF-ORGANIZING MAPS, LEARNING VECTOR QUANTIZATION, INTERPRETABLE MACHINE LEARNING, AND BEYOND, WSOM PLUS 2024, 2024, 1087 : 86 - 95
  • [49] Data minimization for GDPR compliance in machine learning models
    Abigail Goldsteen
    Gilad Ezov
    Ron Shmelkin
    Micha Moffie
    Ariel Farkash
    AI and Ethics, 2022, 2 (3): : 477 - 491
  • [50] Models for machine learning and data mining in functional programming
    Allison, L
    JOURNAL OF FUNCTIONAL PROGRAMMING, 2005, 15 : 15 - 32