Adversarial Information Retrieval

1
Adversarial Information Retrieval

• The Manipulation of Web Content

2
Introduction

• Examples
• TrustRank and Other Methods

3
What is Adversarial IR?

• Gathering, Indexing, Retrieving and Ranking
  Information
• Subset of the information has been manipulated
  maliciously
• Financial Gain

4
What is the Goal of AIR?

• Detect the bad sites or communities
• Improve precision on search engines by
  eliminating the bad guys

5
Simplest form

• First generation engines relied heavily on tf/idf
  
• The top-ranked pages for the query maui resort
  were the ones containing the most mauis and
  resorts
• SEOs responded with dense repetitions of chosen
  terms
• e.g., maui resort maui resort maui resort
• Often, the repetitions would be in the same color
  as the background of the web page
• Repeated terms got indexed by crawlers
• But not visible to humans on browsers

Pure word density cannot be trusted as an IR
signal
6
Search Engine Spamming

• Link-spam
• Link-bombing
• Spam Blogs
• Comment Spam
• Keyword Spam
• Malicious Tagging

7
Spamming

• Online tutorials for search engine persuasion
  techniques
• How to boost your PageRank
• Artificial links and Web communities
• Latest trend Google bombing
• a community of people create (genuine) links with
  a specific anchor text towards a specific page.
  Usually to make a political point

8
Google Bombing
9
Our Focus

• Link Manipulation

10
Trust Rank

• Observation
• Good pages tend to link good pages.
• Human is the best spam detector
• Algorithm
• Select a small subset of pages and let a human
  classify them
• Propagate goodness of pages

11
Propagation

• Trust function T
• T(p) returns the propability that p is a good
  page
• Initial values
• T(p) 1, if p was found to be a good page
• T(p) 0, if p was found to be a spam page
• Iterations
• propagate Trust following out-links
• only a fixed number of iteration M.

12
Propagation (2)

• Problem with propagation
• Pages reachable from good seeds might not be good
• the further away we are from good seed pages, the
  less certain we are that a page is good.

• solution reduce trust as we move further away
  from the good seed pages (trust attenuation).

13
Trust attenuation dampening

• Propagate a dampened trust score ß lt 1 at first
  step
• At n-th step propagate a trust of ßn

14
Trust attenuation splitting

• Parent trust value is splittet among child nodes
• Observation the more the links the less the care
  in choosing them
• Mix damp and split? ßn(splitted trust)

15
Selection Inverse PageRank

• The seed set S should
• be as small as possible
• cover a large part of the Web
• Covering is related to out-links in the very same
  way PageRank is related to in-link
• Inverse PageRank !
• Perform PageRank on a graph with inverted links
• G' (V, E') where (p,q) ? E' ??(q, p) ? E.

16
Algorithm

• Select seeds ( s ) and order by preference
• Invoke oracle (human) on the first L seeds,
• Initialize and normalize oracle response d
• Compute TrustRank score (as in PageRank
  formula) t ß Tt(1-ß) d
• T is the adjacency matrix of the Web Graph.
• ß is the dampening factor. (usually .85)

17
Algorithm - example

• s 0.08, 0.13, 0.08, 0.10, 0.09, 0.06, 0.02
• Ordering 2, 4, 5, 1, 3, 6, 7
• L3 2, 4, 5 d0, 0.5, 0, 0.5, 0, 0, 0
• ß0.85 M20
• t 0, 0.18, 0.12, 0.15, 0.13, 0.05, 0.05
• NB. max0.18
• Issues with page 1 and 5

18
Issues with TrustRank

• Coverage of the seed set may not be broad enough
• Many different topics exist, each with good pages
• TrustRank has a bias towards communities that are
  heavily represented in the seed set
• inadvertently helps spammers that fool these
  communities

19
Bias towards larger partitions

• Divide the seed set into n partitions, each has
  mi nodes
• ti TrustRank score calculated by using
  partition i as the seed set
• t TrustRank score calculated by using all the
  partitions as one combined seed set

20
Basic ideas

• Use pages labeled with topics as seed pages
• Pages listed in highly regarded topic directories
• Trust should be propagated by topics
• link between two pages is usually created in a
  topic specific context

21
Topical TrustRank

• Topical TrustRank
• Partition the seed set into topically coherent
  groups
• TrustRank is calculated for each topic
• Final ranking is generated by a combination of
  these topic specific trust scores
• Note
• TrustRank is essentially biased PageRank
• Topical TrustRank is fundamentally the same as
  Topic-Sensitive PageRank, but for demoting spam

22
Combination of trust scores

• Simple summation
• default mechanism just seen
• Quality bias
• Each topic weighted by a bias factor
• Summation of these weighted topic scores
• One possible bias Average PageRank value of the
  seed pages of the topic

23
Further Improvements

• Seed Weighting
• Instead of assigning an equal weight to each seed
  page, assign a weight proportional to its quality
  / importance
• Seed Filtering
• Filtering out low quality pages that may exist in
  topic directories
• Finer topics
• Lower layers of the topic directory