| Paper Abstract and Keywords |
| Presentation |
2015-11-26 15:00
[Poster Presentation]
Theoretical Vulnerability Evaluation on Linear Classifier under Self-Information Controllable Settings Shohei Kobayashi (Univ. of Tsukuba), Shota Okumura, Ichiro Takeuchi (nitech), Jun Sakuma (Univ. of Tsukuba) IBISML2015-64 |
| Abstract |
(in Japanese) |
(See Japanese page) |
| (in English) |
Self-information controllable learning is machine learning algorithms that allow data providers to control their data even after contributing their data to the learner. This setting provides proactive privacy protection for data providers. Meanwhile, if this system is misused, the training data can be corrupted by user account hijacking or injection of poisonous data by shilling users. This might degrade prediction accuracy of machine learning. In this manuscript, we formulate the attack against self-information controllable linear classification problem. We define the vulnerability of the learning algorithm as possible changes in prediction caused by data modification. The vulnerability can be theoretically evaluated by quick sensitivity analysis, which evaluates the changes in the model caused by addition, removal, and modification of training examples. Our analysis revealed that the expected change of prediction is proportional to $O(m/n)$, where $m$ is the number of modified data and $n$ is the number of the training data. The result demonstrates that, given a large number of training data, the adversary can degrade the classification accuracy only by increasing the total number of the data to be modified. Also, we empirically confirmed the theoretical result with a real dataset. |
| Keyword |
(in Japanese) |
(See Japanese page) |
| (in English) |
Self-Information controllable / linear classification / vulnerability evaluation / / / / / |
| Reference Info. |
IEICE Tech. Rep., vol. 115, no. 323, IBISML2015-64, pp. 83-90, Nov. 2015. |
| Paper # |
IBISML2015-64 |
| Date of Issue |
2015-11-19 (IBISML) |
| ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
| Download PDF |
IBISML2015-64 |
| Conference Information |
| Committee |
IBISML |
| Conference Date |
2015-11-25 - 2015-11-27 |
| Place (in Japanese) |
(See Japanese page) |
| Place (in English) |
Epochal Tsukuba |
| Topics (in Japanese) |
(See Japanese page) |
| Topics (in English) |
Information-Based Induction Science Workshop (IBIS2015) |
| Paper Information |
| Registration To |
IBISML |
| Conference Code |
2015-11-IBISML |
| Language |
Japanese |
| Title (in Japanese) |
(See Japanese page) |
| Sub Title (in Japanese) |
(See Japanese page) |
| Title (in English) |
Theoretical Vulnerability Evaluation on Linear Classifier under Self-Information Controllable Settings |
| Sub Title (in English) |
|
| Keyword(1) |
Self-Information controllable |
| Keyword(2) |
linear classification |
| Keyword(3) |
vulnerability evaluation |
| Keyword(4) |
|
| Keyword(5) |
|
| Keyword(6) |
|
| Keyword(7) |
|
| Keyword(8) |
|
| 1st Author's Name |
Shohei Kobayashi |
| 1st Author's Affiliation |
University of Tsukuba (Univ. of Tsukuba) |
| 2nd Author's Name |
Shota Okumura |
| 2nd Author's Affiliation |
Nagoya Institute of Technology (nitech) |
| 3rd Author's Name |
Ichiro Takeuchi |
| 3rd Author's Affiliation |
Nagoya Institute of Technology (nitech) |
| 4th Author's Name |
Jun Sakuma |
| 4th Author's Affiliation |
University of Tsukuba (Univ. of Tsukuba) |
| 5th Author's Name |
|
| 5th Author's Affiliation |
() |
| 6th Author's Name |
|
| 6th Author's Affiliation |
() |
| 7th Author's Name |
|
| 7th Author's Affiliation |
() |
| 8th Author's Name |
|
| 8th Author's Affiliation |
() |
| 9th Author's Name |
|
| 9th Author's Affiliation |
() |
| 10th Author's Name |
|
| 10th Author's Affiliation |
() |
| 11th Author's Name |
|
| 11th Author's Affiliation |
() |
| 12th Author's Name |
|
| 12th Author's Affiliation |
() |
| 13th Author's Name |
|
| 13th Author's Affiliation |
() |
| 14th Author's Name |
|
| 14th Author's Affiliation |
() |
| 15th Author's Name |
|
| 15th Author's Affiliation |
() |
| 16th Author's Name |
|
| 16th Author's Affiliation |
() |
| 17th Author's Name |
|
| 17th Author's Affiliation |
() |
| 18th Author's Name |
|
| 18th Author's Affiliation |
() |
| 19th Author's Name |
|
| 19th Author's Affiliation |
() |
| 20th Author's Name |
|
| 20th Author's Affiliation |
() |
| 21st Author's Name |
|
| 21st Author's Affiliation |
() |
| 22nd Author's Name |
|
| 22nd Author's Affiliation |
() |
| 23rd Author's Name |
|
| 23rd Author's Affiliation |
() |
| 24th Author's Name |
|
| 24th Author's Affiliation |
() |
| 25th Author's Name |
|
| 25th Author's Affiliation |
() |
| 26th Author's Name |
/ / |
| 26th Author's Affiliation |
()
() |
| 27th Author's Name |
/ / |
| 27th Author's Affiliation |
()
() |
| 28th Author's Name |
/ / |
| 28th Author's Affiliation |
()
() |
| 29th Author's Name |
/ / |
| 29th Author's Affiliation |
()
() |
| 30th Author's Name |
/ / |
| 30th Author's Affiliation |
()
() |
| 31st Author's Name |
/ / |
| 31st Author's Affiliation |
()
() |
| 32nd Author's Name |
/ / |
| 32nd Author's Affiliation |
()
() |
| 33rd Author's Name |
/ / |
| 33rd Author's Affiliation |
()
() |
| 34th Author's Name |
/ / |
| 34th Author's Affiliation |
()
() |
| 35th Author's Name |
/ / |
| 35th Author's Affiliation |
()
() |
| 36th Author's Name |
/ / |
| 36th Author's Affiliation |
()
() |
| Speaker |
Author-1 |
| Date Time |
2015-11-26 15:00:00 |
| Presentation Time |
180 minutes |
| Registration for |
IBISML |
| Paper # |
IBISML2015-64 |
| Volume (vol) |
vol.115 |
| Number (no) |
no.323 |
| Page |
pp.83-90 |
| #Pages |
8 |
| Date of Issue |
2015-11-19 (IBISML) |