"Book Notes" recommendation System (Recommender systems An Introduction) Nineth chapter on the attack of collaborative filtering recommendation system

Source: Internet
Author: User

Personal evaluation: A very interesting topic, I also encountered in the actual work, but the general writing, a little "Shini", too academic, too yy, the premise is too strong. Let's take a look at it for reference.
It is generally recommended that the system use user data when it is assumed that the user is kind and honest. While attacking, the only value is to think of trying to influence the system's results, performance.
Dimension of the attack: 1. Raise or lower the score for an item, 2. For a specific user base; 3. For a system, the system recommendation is inaccurate, or even the system crashes.
All attacks are done in some way (for example, simulating user behavior) to inject some specific data into the recommender system to achieve the goal.
Attack Type: 1. Random attacks inject random values into the system to achieve the purpose of interfering with the recommended system. 2. Mean attack on the basis of random attacks, using the scoring mean, to construct more "like" injection data 3. What makes a build-up attack more sophisticated than a mean attack is that, in addition to the high scoring (or low split) of the target item, there is a high score for many popular items, with the goal that many users ' usage records also contain these popular items, and that the injected data is easier to form a "neighbor" relationship with the user. This makes it easier to adopt the recommended system, which is more likely to affect the end user. 4. A local attack is more sophisticated than a campaign attack in that a local attack can identify a particular user group and attack it accordingly. Personal evaluation: Identifying specific user groups should be difficult in an attack, unless some data from the Recommender system is stolen. 5. Targeted suppression attacks are different from the goal of raising an item's rating, and suppressing the attack is to reduce the score of an item. Method is the method above, the opposite. In general, the crackdown is more likely to succeed-the academic community has not yet explained why this asymmetry exists. 6. Click Stream attack and implicit feedback means to simulate the user's operation on the Web page to achieve the purpose of injecting data. -Personally suspect, how did the previous five methods inject data? It is difficult to directly modify the background database of others?! Clickstream attacks usually affect the "many of the classmates who read the book read the book."
Attack countermeasure 1. Increase data injection costs by 2. Automatic detection of abnormal data by different systems, such as: In a short time in the same direction large amounts of data entry, compared to other data of abnormal data entry
Protecting data privacy 1. Adding random disturbances to normal data

2. Distributed storage--the same user's data is in multiple places, even if one place leaks, can not get the entire user record


Finish.


"Book Notes" recommendation System (Recommender systems An Introduction) Nineth chapter on the attack of collaborative filtering recommendation system

Contact Us

The content source of this page is from Internet, which doesn't represent Alibaba Cloud's opinion; products and services mentioned on that page don't have any relationship with Alibaba Cloud. If the content of the page makes you feel confusing, please write us an email, we will handle the problem within 5 days after receiving your email.

If you find any instances of plagiarism from the community, please send an email to: info-contact@alibabacloud.com and provide relevant evidence. A staff member will contact you within 5 working days.

A Free Trial That Lets You Build Big!

Start building with 50+ products and up to 12 months usage for Elastic Compute Service

  • Sales Support

    1 on 1 presale consultation

  • After-Sales Support

    24/7 Technical Support 6 Free Tickets per Quarter Faster Response

  • Alibaba Cloud offers highly flexible support services tailored to meet your exact needs.