@article{fdi:010090790,
  title = {{A}dversarial attacks and defenses in person search : a systematic mapping study and taxonomy},
  author = {{A}ndrade, {E}. {D}. and {G}uerin, {J}oris and {V}iterbo, {J}. and {S}ampaio, {I}. {G}. {B}.},
  editor = {},
  language = {{ENG}},
  abstract = {{P}erson {S}earch aims at retrieving a specific individual (the query) within a collection of whole scene images from diverse, non -overlapping cameras. {I}t has the potential to play a pivotal role in various public safety applications like suspect searching and identifying abandoned luggage owners. {P}erson {S}earch encompasses two {C}omputer {V}ision challenges: 1. {O}bject {D}etection, which entails localizing humans in whole scene images, and 2. {P}erson {R}e{I}dentification, where the query image is compared with images of detected individuals to establish identification. {T}he critical nature of {P}erson {S}earch underscores the imperative to safeguard it against security threats, such as adversarial attacks, which can result in non-detection or misidentification. {W}hile adversarial attacks and defense mechanisms have been extensively studied for both {O}bject {D}etection and {P}erson {R}e -{I}dentification, there is a noticeable gap in research concerning {P}erson {S}earch. {T}his work presents a comprehensive {S}ystematic {M}apping {S}tudy and taxonomy of adversarial attacks and defenses in {P}erson {S}earch, utilizing {P}arsifal and {C}hat{GPT} 4 for indepth analysis. {W}e highlight the persistent challenges associated with {P}erson {S}earch and discuss prospects for future advancements in addressing its vulnerabilities.},
  keywords = {{P}erson search ; {P}erson re -identification ; {O}bject detection ; {A}dversarial attacks ; {A}dversarial defenses},
  booktitle = {},
  journal = {{I}mage and {V}ision {C}omputing},
  volume = {148},
  numero = {},
  pages = {105096 [14 ]},
  ISSN = {0262-8856},
  year = {2024},
  DOI = {10.1016/j.imavis.2024.105096},
  URL = {https://www.documentation.ird.fr/hor/fdi:010090790},
}