This document provides solutions for maximizing the PageRank of a website called rolexx.com by adding other pages and strategically linking to them.
There are two cases:
Case a) adds k pages, links them all to rolexx.com, and links rolexx.com to one of the k pages. This maximizes the probability of visiting rolexx.com.
Case b) uses the same structure but also links all legitimate websites to rolexx.com to further increase the chances of visiting it. This allows rolexx.com to be reached in two hops.
Problem 2 notes that the PageRank calculation can account for different link weights by making the transition probabilities proportional to the link
This document provides solutions for maximizing the PageRank of a website called rolexx.com by adding other pages and strategically linking to them.
There are two cases:
Case a) adds k pages, links them all to rolexx.com, and links rolexx.com to one of the k pages. This maximizes the probability of visiting rolexx.com.
Case b) uses the same structure but also links all legitimate websites to rolexx.com to further increase the chances of visiting it. This allows rolexx.com to be reached in two hops.
Problem 2 notes that the PageRank calculation can account for different link weights by making the transition probabilities proportional to the link
This document provides solutions for maximizing the PageRank of a website called rolexx.com by adding other pages and strategically linking to them.
There are two cases:
Case a) adds k pages, links them all to rolexx.com, and links rolexx.com to one of the k pages. This maximizes the probability of visiting rolexx.com.
Case b) uses the same structure but also links all legitimate websites to rolexx.com to further increase the chances of visiting it. This allows rolexx.com to be reached in two hops.
Problem 2 notes that the PageRank calculation can account for different link weights by making the transition probabilities proportional to the link
This document provides solutions for maximizing the PageRank of a website called rolexx.com by adding other pages and strategically linking to them.
There are two cases:
Case a) adds k pages, links them all to rolexx.com, and links rolexx.com to one of the k pages. This maximizes the probability of visiting rolexx.com.
Case b) uses the same structure but also links all legitimate websites to rolexx.com to further increase the chances of visiting it. This allows rolexx.com to be reached in two hops.
Problem 2 notes that the PageRank calculation can account for different link weights by making the transition probabilities proportional to the link
1 Problem 1 The goal is to find how many pages of the k pages we need to add and how we should interlink them to maximize the PageRank of rolexx.com. Lets assume that we have some pages already interlinked by the spammer and see what happens if the random surfer is walking around. The random surfer can be modelled as having three states:
R - visiting the rolexx.com website
K - visiting one of the l < k websites added by the spammer
O - visiting the other (legitimate) websites
Since the crawler has no memory, we can model this as a three-state Markov chain (Figure 1. Maximizing the PageRank of rolexx.com is equivalent to maximizing the probability that rolexx.com is visited i.e. the probability of being in state R (pR ). We have pR = pK pKR + pO pOR + pR pRR , which transforms to pR = pK pKR +pO pOR 1pRR and this is the expression that we want to maximize. 1 pRR is a constant so we ignore it and only need to maximize pR = pK pKR + pO pOR Case a.) To maximize pKR we need to link from all the l pages to rolexx.com and put no links among the l pages (pKK minimized) and from the l pages to other pages (pKO minimized). We also set l = k to maximize pK . If we dont put any links from rolexx.com to anywhere then it becomes a sink and the random
pOO pRR pRO
pOR
pKO pKR pOK pRK
pKK
Figure 1: The Markov state transition diagram for the surfer.
1 N legitimate pages
rolexx.com
other pages added
by the spammer
Figure 2: Solution for case a.)
N legitimate pages
rolexx.com
other pages added
by the spammer
Figure 3: Solution for case b.)
walker has no link to take next and will always select an arbitrary page and jump to it. To prevent that we link from rolexx.com to one of the k pages. The final interlinking that maximizes the PageRank of rolexx.com in the a.) case is on Figure 2. Case b.) We keep the link structure from a.) but now we can additionally control the pOR . To maximize this probability we link from all the legitimate websites to rolexx.com. We can also increase the chances of visiting rolexx.com by linking via one of the k websites, then rolexx.com can be reached in two hops by the surfer. We create a link from all legitimate websites to the k websites. The final interlinking is on Figure 3. The spamming in case a.) is feasible to do in the context of Google, the spammer can easily create an arbitrary number of new websites and link between them. In case b.) it is a bit hard to convince the whole World Wide Web to link to the k + 1 pages. Please have a look at the Wikipedia entry for link farms for a discussion.
2 2 Problem 2 Lets assume that the weight of the link from the page pi to pj is wij . The weight equals zero if there is no link. The only modification that needs to be made is wij to set the values of the transition matrix Rij = k=1..n wik which is simply the weights normalized by their sum to have transition probabilities proportional to the weights.
