BEGIN:VCALENDAR
VERSION:2.0
PRODID:Data::ICal 0.24
BEGIN:VEVENT
DESCRIPTION:   'Title: QDoor: Exploiting Approximate Synthesis for Backdoor
  Attacks\n   in Quantum Neural Networks\n   When: Saturday\, Aug 12\, 13:0
 0 - 13:59 PDT\n   Where: LINQ - 3rd flr - Quantum Village - [1]Map\n\n   S
 peakerBio:Lei Jiang \, Assoc. Prof. at Indiana University Bloomington\n   
 No BIO available\n\n   Description:\n   Quantum neural networks (QNNs) suc
 ceed in object recognition\, natural\n   language processing\, and financi
 al analysis. To maximize the accuracy\n   of a QNN on a Noisy Intermediate
  Scale Quantum (NISQ) computer\,\n   approximate synthesis modifies the QN
 N circuit by reducing error-prone\n   2-qubit quantum gates. The success o
 f QNNs motivates adversaries to\n   attack QNNs via backdoors. However\, n
 a¨Ä±vely transplanting backdoors\n   designed for classical neural network
 s to QNNs yields only low attack\n   success rate\, due to the noises and 
 approximate synthesis on NISQ\n   computers. Prior quantum circuit-based b
 ackdoors cannot selectively\n   attack some inputs or work with all types 
 of encoding layers of a QNN\n   circuit. Moreover\, it is easy to detect b
 oth transplanted and\n   circuit-based backdoors in a QNN.\n\n   In this t
 alk\, we introduce a novel and stealthy backdoor attack\,\n   QDoor\, to a
 chieve high attack success rate in\n   approximately-synthesized QNN circu
 its by weaponizing unitary\n   differences between uncompiled QNNs and the
 ir synthesized\n   counterparts. QDoor trains a QNN behaving normally for 
 all inputs with\n   and without a trigger. However\, after approximate syn
 thesis\, the QNN\n   circuit always predicts any inputs with a trigger to 
 a predefined\n   class while still acts normally for benign inputs. Compar
 ed to prior\n   backdoor attacks\, QDoor improves the attack success rate 
 by 13× and\n   the clean data accuracy by 65% on average. Furthermore\, pr
 ior backdoor\n   detection techniques cannot find QDoor attacks in uncompi
 led QNN\n   circuits.\n\n   '\n\n   1. #Linq\n\n\n
DTEND:20230812T205900Z
DTSTART:20230812T200000Z
LOCATION:QTV - LINQ - 3rd flr - Quantum Village
SUMMARY:QDoor: Exploiting Approximate Synthesis for Backdoor Attacks in Qua
 ntum Neural Networks
END:VEVENT
END:VCALENDAR