CS2125 Paper Review Form - Winter 2019

Reviewer: Ali Harakeh

Paper Title: Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey

Author(s): Akhtar and Mian

1) Is the paper technically correct?
 [X] Yes
 [ ] Mostly (minor flaws, but mostly solid)
 [ ] No

2) Originality
 [ ] Very good (very novel, trailblazing work)
 [ ] Good 
 [ ] Marginal (very incremental)
 [X] Poor (little or nothing that is new)

3) Technical Depth
 [ ] Very good (comparable to best conference papers)
 [ ] Good (comparable to typical conference papers)
 [X] Marginal depth
 [ ] Little or no depth

4) Impact/Significance
 [ ] Very significant
 [X] Significant
 [ ] Marginal significance.
 [ ] Little or no significance.

5) Presentation
 [ ] Very well written
 [X] Generally well written
 [ ] Readable
 [ ] Needs considerable work
 [ ] Unacceptably bad

6) Overall Rating
 [ ] Strong accept (award quality)
 [X] Accept (high quality - would argue for acceptance)
 [ ] Weak Accept (borderline, but lean towards acceptance)
 [ ] Weak Reject (not sure why this paper was published)


7) Summary of the paper's main contribution and rationale
   for your recommendation. (1-2 paragraphs)
This paper presents the first comprehensive survey on adversarial attacks, 
the possibility of their existence, and proposed defenses against them. 
The paper is quite informative and dense, and covers the most important milestones on adversarial attacks 
as well as the state of the art. In addition, the survey sheds light on the most important aspects 
of possible future research on this topic.

I am not an expert in this topic, but up to my knowledge, the survey is quite comprehensive.

8) List 1-3 strengths of the paper.  (1-2 sentences each,
identified as S1, S2, S3.)

S1: The best part of the paper is the final outlook on research directions. Conclusions state in that sections are very good directions for a researcher looking to contribute to adversarial deep learning.

S2: The comprehensiveness of this survey makes it a prime target as a starter paper for researchers new to adversarial deep learning.

S3: Survey papers tend to fall into being too general. This paper on the other hand maintains focus all throughout, with very little diversions making it an easy read.  

9) List 1-3 weaknesses of the paper (1-2 sentences each,
identified as W1, W2, W3.)

W1: Very little quantitative comparison is provided in the survey sections, making it of little use to expert audience. Usually, a survey would provide some guidance on which attack is to be used for testing which characteristics of a neural network and so on.