blatyo/trust_gaurd

Implementation of the algorithm found in TrustGaurd: Countering Vulnerabilities in Reputation Management for Decentralized Overlay Networks.

TrustGuard

This is my implementation of the algorithm described in TrustGaurd: Countering Vulnerabilities in Reputation Management for Decentralized Overlay Networks.. I haven't tested it. I just wanted to write it in code so that I could figure out what each piece meant.

Also, I realize I misspelled guard as gaurd everywhere. Its just one of those words I misspell all the time and its not worth fixing.

Usage

require 'trust_gaurd'

# current raw trust value gets 2/3rds of the importance
alpha = 0.66

# previous raw trust values get 1/3rd of the importance
beta = 0.34

# a positive change in reputation rewards double the change
pos_gamma = 2

# a negative change in reputation punishes four times the change
neg_gamma = 4

# the importance of previous raw trust values degrades by a factor of 0.5
p = 0.5

tv = TrustGaurd::TrustValue.new(alpha, beta, pos_gamma, neg_gamma, p)

# Gets all of raw trust value, nothing from history, and no change
tv.calculate([100])                   #=> 66.0
# Gets all of raw trust value, all of history, and no change
tv.calculate([100, 100])              #=> 100.0
# Gets 95% of raw trust value, all of history, and -5 * 4 change
tv.calculate([100, 100, 95])          #=> 76.70000000000006
# The rest gets a bit complicated...
tv.calculate([100, 100, 95, 95])      #=> 87.15714285714287
tv.calculate([100, 100, 95, 95, 77])  #=> 7.460000000000008

Variables

Some notes about varying these values:

• When alpha > beta, history does not have as much significance as the current value. The opposite is also true.
• When pos_gamma is large, it is easier to obtain a higher trust value over time. The authors that propose this algorithm suggest that it be hard to get a good reputation.
• When pos_gamma is large, there is a large penalty for a bad reputation. The authors that propose this algorithm suggest that reputation be lost quickly if the raw trust history decreases radically.
• When p = 1, all previous raw trust values are treated equally. Generally, more recent ones should have more significance. Therefore, a p value < 1 is preferable.

Note on Patches/Pull Requests

• Fork the project.
• Make your feature addition or bug fix.
• Add tests for it. This is important so I don't break it in a future version unintentionally.
• Commit, do not mess with rakefile, version, or history. (if you want to have your own version, that is fine but bump version in a commit by itself I can ignore when I pull)
• Send me a pull request. Bonus points for topic branches.

Copyright

Copyright (c) 2010 Allen Madsen. See LICENSE for details.