Issues

• What's the difference between UMIFGSM and TMIFGSM

  ryderling/DEEPSEC#16 · Filter by author XiaodanLi001 opened on Jul 16, 2019
  2

• PGD/BIM implementation is incorrect

  ryderling/DEEPSEC#15 · Filter by author carlini opened on Mar 24, 2019
  1

• JSMA implementation is incorrect

  ryderling/DEEPSEC#14 · Filter by author carlini opened on Mar 24, 2019

• Significant and fundamental flaws in methodology, analysis, and conclusions

  ryderling/DEEPSEC#13 · Filter by author carlini opened on Feb 26, 2019
  1

• Discrepancies between tables, text, and code

  ryderling/DEEPSEC#12 · Filter by author carlini opened on Feb 26, 2019
  7

• Paper does not report attack success rate for targeted adversarial examples

  ryderling/DEEPSEC#11 · Filter by author carlini opened on Feb 26, 2019
  2

• Reporting success rate of unbounded attacks is meaningless

  ryderling/DEEPSEC#10 · Filter by author carlini opened on Feb 26, 2019
  2

• Attack success rate decreases with distortion bound

  ryderling/DEEPSEC#9 · Filter by author carlini opened on Feb 26, 2019
  2

• Epsilon values studied are too large to be meaningful

  ryderling/DEEPSEC#7 · Filter by author carlini opened on Feb 26, 2019
  4

• Comparing attack effectiveness is done incorrectly

  ryderling/DEEPSEC#6 · Filter by author carlini opened on Feb 26, 2019
  2

• Computing the average over different threat models is meaningless

  ryderling/DEEPSEC#5 · Filter by author carlini opened on Feb 26, 2019
  2

• PGD adversarial training implementation is incorrect

  ryderling/DEEPSEC#4 · Filter by author carlini opened on Feb 26, 2019
  3

• FGSM implementation is incorrect

  ryderling/DEEPSEC#3 · Filter by author carlini opened on Feb 26, 2019
  9

• Paper uses averages instead of the minimum for security analysis

  ryderling/DEEPSEC#2 · Filter by author carlini opened on Feb 26, 2019
  2

• Attacks are not run on defenses in an all-pairs manner

  ryderling/DEEPSEC#1 · Filter by author carlini opened on Feb 26, 2019
  2