PhD Seminar • Software Testing | Artificial Intelligence • Finding Deviated Behaviors of the Compressed DNN Models for Image ClassificationsExport this event to calendar

Friday, March 10, 2023 — 1:00 PM to 2:00 PM EST

Please note: This PhD seminar will take place in DC 2564.

Yongqiang (Victor) Tian, PhD candidate
David R. Cheriton School of Computer Science

Supervisor: Professor Chengnian Sun

Model compression can significantly reduce the sizes of deep neural network (DNN) models, and thus facilitates the dissemination of sophisticated, sizable DNN models, especially for their deployment on mobile or embedded devices. However, the prediction results of compressed models may deviate from those of their original models. To help developers thoroughly understand the impact of model compression, it is essential to test these models to find those deviated behaviors before dissemination. However, this is a non-trivial task because the architectures and gradients of compressed models are usually not available.

To this end, we propose Dflare, a novel, search-based, black-box testing technique to automatically find triggering inputs that result in deviated behaviors in image classification tasks. Dflare iteratively applies a series of mutation operations to a given seed image, until a triggering input is found. For better efficacy and efficiency, Dflare models the search problem as Markov Chains and leverages the Metropolis-Hasting algorithm to guide the selection of mutation operators in each iteration. Further, Dflare utilizes a novel fitness function to prioritize the mutated inputs that either cause large differences between two models’ outputs, or trigger previously unobserved models’ probability vectors.

We evaluated Dflare on 21 compressed models for image classification tasks with three datasets. The results show that Dflare not only constantly outperforms the baseline in terms of efficacy, but also significantly improves the efficiency: Dflare is 17.84x ∼ 446.06x as fast as the baseline in terms of time; the number of queries required by Dflare to find one triggering input is only 0.186% ∼ 1.937% of those issued by the baseline. We also demonstrated that the triggering inputs found by Dflare can be used to repair up to 48.48 deviated behaviors in image classification tasks and further decrease the effectiveness of Dflare on the repaired models.

Location 
DC - William G. Davis Computer Research Centre
DC 2564
200 University Avenue West

Waterloo, ON N2L 3G1
Canada
Event tags 

S M T W T F S
25
26
27
28
29
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
1
2
3
4
5
6
  1. 2024 (100)
    1. April (23)
    2. March (27)
    3. February (25)
    4. January (25)
  2. 2023 (296)
    1. December (20)
    2. November (28)
    3. October (15)
    4. September (25)
    5. August (30)
    6. July (30)
    7. June (22)
    8. May (23)
    9. April (32)
    10. March (31)
    11. February (18)
    12. January (22)
  3. 2022 (245)
  4. 2021 (210)
  5. 2020 (217)
  6. 2019 (255)
  7. 2018 (217)
  8. 2017 (36)
  9. 2016 (21)
  10. 2015 (36)
  11. 2014 (33)
  12. 2013 (23)
  13. 2012 (4)
  14. 2011 (1)
  15. 2010 (1)
  16. 2009 (1)
  17. 2008 (1)