Resistance to Sybil and otherwise Malicious Attacks
First set of controls to learn
Anchor (a.k.a. Seed User):
seed user selector moved to top panel (basicGrapevineControlsTopPanel.php)
The Anchor (Seed) User is the center of your Grapevine. By definition, the Influence of the Anchor User is 1.
You can change Anchor User whenever you want, however many times you want.
The default Anchor is OTB-Default, controlled by OnTheBlockchain.
Once you have provided Trust Ratings for one or more other users, it is worthwhile to make yourself the Anchor User.
You may also set any other user of your choosing to be the Anchor [work in progress].
Decouple Rigor from Confidence in Default User Trust Score:
no, leave coupled
yes, uncouple
Currently if either the Rigor or the Default Confidence (User Trust) is changed, then the other variable is also changed in a manner so that for an unvetted user,
the Confidence (set by the user) and the Certainty (calculated using the Rigor Equation) are equal. If Confidence goes up, Rigor goes down, and vice versa.
Rigor
%
Rigor is a parameter in the equation that controls the relationship between the Input (a.k.a. "Work") and the Certainty of any composite score.
Currently, the value for Rigor is the same for every composite score, although we may make it possible in the future to use different values for different composite scores.
Certainty (%) = 100 * (1 - e-Input*ln(Rigor) )
USER TRUST SCORE
brief description of the user trust score
Most important component:
Influence
Average Score
?
This is the component of the User Trust Composite Score that is frequently displayed next to the user name.
Typically the "Influence" component is chosen for user trust scores, but you may wish to switch it to "Average Score" if that makes more sense to you.
This option does not change how any scores get calculated; only how they are displayed.
Second important component:
Input
Certainty
Number of Ratings
?
This is the component of the User Trust Composite Score that is frequently displayed next to the user name.
Typically the "Influence" component is chosen for user trust scores, but you may wish to switch it to "Average Score" if that makes more sense to you.
This option does not change how any scores get calculated; only how they are displayed.
default user sliders moved to top panel (basicGrapevineControlsTopPanel.php)
Weight adjustment:
off (no adjustment)
standard
special algorithm
LISTINGS: DEFAULT RATINGS (5-Star)
Average score on a scale of 1 to 5.
Most important component:
Influence
Average Score
?
This is the component of the Listing Score that is displayed next to the listing.
Typically the "Average Score" component is chosen for listing scores, but you may wish to switch it to "Influence" which has the effect of stratifying listings based on a combination of average score and amount of input.
This option does not change how any scores get calculated; only how they are displayed.
Second most important component:
Input
Certainty
Number of Ratings
?
This is the component of the Listing Score that is displayed next to the listing.
Typically the "Average Score" component is chosen for listing scores, but you may wish to switch it to "Influence" which has the effect of stratifying listings based on a combination of average score and amount of input.
This option does not change how any scores get calculated; only how they are displayed.
Sliders for default listing scores has been moved to Listing Table tab
Weight adjustment:
off (no adjustment)
standard
special algorithm
MODERATOR TRUST SCORE
The Moderator Trust Score is derived from several inputs which are shown below.
The contributions of each input can be adjusted using the relevant Attenuation Factor, controlled by scrollbars, below.
Setting any given input's Attenuation Factor to zero effectively removes that input.
In addition, the contributions of "Default" and "Inherited" inputs are gradually reduced to zero as contributions accumulate from more specific inputs (the "Flag-Type" inputs).
This is described as "Weight Adjustment" and can be turned on or off using selector elements below.
Most important component:
Influence
Average Score
?
This is the component of the Moderator Score that is frequently displayed next to the Moderator name.
The default option is for the "Influence Score" component to be chosen for Moderator Scores, but you may wish to switch it to "Average Score" if that makes more sense to you.
This option does not change how any scores get calculated; only how they are displayed (and perhaps stratified).
Second most important component:
Input
Certainty
Number of Ratings
?
This is the component of the score that is frequently displayed in parentheses next to the main score to indicate how much effort went into its calculation.
Legacy ratings systems give number of ratings as the only option, as there is no concept of Input or Certainty.
Default Score
default score
Your default assignment of the comp score of an un-vetted Moderator. Red flags => -1000, green flags ==> +1000
confidence:
%
How much of you stands behind that guess?
Weight adjustment:
off (no adjustment)
standard
special algorithm
?
For Default Scores, Weight Adjustment means subtracting the weight of the Default Score from the total weights of all non-inherited scores.
The effect is that the default settings kicks in when there is no other, more relevant information, but is gradually phased out as more relevant information is obtained.
"Inherited" Trust
Composite Score:
none
Generic User Trust
something else
exclude defaults?
no
yes
Rating =
Influence
Average Score
Weight =
Certainty
Input
* attenuation factor (AF)
AF:
OFF
ON
Weight adjustment:
off (no adjustment)
standard
special algorithm
?
For an Inherited Score, Weight Adjustment means subtracting the weight of the Inherited Score from the total weights of all non-inherited scores.
The effect is that Inheritance kicks in when there is no other, more relevant information, but is gradually phased out as more relevant information is obtained.
FLAG-TYPE INPUTS
Composite Score for: Flag Moderator: OB1 Verified
Rating =
+1
0
-1
Weight =
Certainty
Input
* attenuation factor (AF)
AF:
OFF
ON
Composite Score for: Flag Moderator: Legit
Rating =
+1
0
-1
Weight =
Certainty
Input
* attenuation factor (AF)
AF:
OFF
ON
Composite Score for: Flag Moderator: Scam
Rating =
+1
0
-1
Weight =
Certainty
Input
* attenuation factor (AF)
AF:
OFF
ON
Composite Score for: Flag Moderator: Test
Rating =
+1
0
-1
Weight =
Certainty
Input
* attenuation factor (AF)
AF:
OFF
ON
VENDOR TRUST SCORE
The Vendor Trust Score is a weighted average of several composite scores.
The majority of the input comes from Composite Scores from several individual "Flag" ratings (one Composite Score for each type of Flag), in which users flag Vendors Vendors in one of several ways: Scam, Illegal, NSFW, Test (which are a score of -1) or Legit (which is a score of +1).
If desired, the relative weight of the individual flag-type Composite Scores can be adjusted by adjusting the associated Attenuation Factors.
In addition, the Vendor Trust Score "inherits" a score from the Generic User Trust. In this way, Vendors may be trusted by virtue of being generically trusted users, even if there are few or no flag ratings.
This may be of benefit to a user who is newly setting up a store or only sells a small number of items. However, the impact of inheritance may be decreased or set to zero by decreasing the associated Attenuation Factor.
(Future: allow "inheritance" from other scores; allow inheritance from multiple scores simultaneously; all to change the rating continuously; allow to change Weight to Input.)
V
Vendor Trust Score: Display
The Vendor Trust Score is typically displayed with the following symbol:
V
There are two components. The first component is the "Vendor Score" itself and may be used as a visual aide or to stratify search results.
The second component is an indicator of how much faith you have in the calculation of the first score. When displayed, it is often in parentheses.
Below, choose what to use throughout the rest of the site as the primary and secondary components of the Vendor Trust Score:
Most important component:
Influence
Average Score
?
This is the component of the Vendor Score that is frequently displayed next to the Vendor name.
The default option is for the "Influence Score" component to be chosen for Vendor Scores, but you may wish to switch it to "Average Score" if that makes more sense to you.
This option does not change how any scores get calculated; only how they are displayed.
Second most important component:
Input
Certainty
Number of Ratings
?
This is the component of the score that is frequently displayed in parentheses next to the main score to indicate how much effort went into its calculation.
Legacy ratings systems give number of ratings as the only option, as there is no concept of Input or Certainty.
Default Score
default score
Your default assignment of the Average Score of an un-vetted vendor.
confidence:
%
How much of you stands behind that guess?
Weight adjustment:
off (no adjustment)
standard
special algorithm
?
For Default Scores, Weight Adjustment means subtracting the weight of the Default Score from the total weights of all non-inherited scores.
The effect is that the default settings kicks in when there is no other, more relevant information, but is gradually phased out as more relevant information is obtained.
"Inherited" Trust
Composite Score:
none
Generic User Trust
something else
exclude defaults?
no
yes
Rating =
Influence
Average Score
Weight =
Certainty
Input
* attenuation factor (AF)
AF:
OFF
ON
Weight adjustment:
off (no adjustment)
standard
special algorithm
?
For an Inherited Score, Weight Adjustment means subtracting the weight of the Inherited Score from the total weights of all non-inherited scores.
The effect is that Inheritance kicks in when there is no other, more relevant information, but is gradually phased out as more relevant information is obtained.
FLAG-TYPE INPUTS
Composite Score for: Flag Vendor: Legit
Rating =
+1
0
-1
Weight =
Certainty
Input
* attenuation factor (AF)
AF:
OFF
ON
Composite Score for: Flag Vendor: Scam
Rating =
+1
0
-1
Weight =
Certainty
Input
* attenuation factor (AF)
AF:
OFF
ON
Composite Score for: Flag Vendor: Illegal
Rating =
+1
0
-1
Weight =
Certainty
Input
* attenuation factor (AF)
AF:
OFF
ON
Composite Score for: Flag Vendor: NSFW
Rating =
+1
0
-1
Weight =
Certainty
Input
* attenuation factor (AF)
AF:
OFF
ON
Composite Score for: Flag Vendor: Test
Rating =
+1
0
-1
Weight =
Certainty
Input
* attenuation factor (AF)
AF:
OFF
ON
Overview
Adjust Defaults
Att. Factor
Mod #1
Mod #2
Mod #3
Bad actors, to effect their nefarious schemes, will attempt to increase the INFLUENCE of one or more user profiles under their control by rating themselves highly and / or providing low trust ratings to other users who have correctly identified their profiles as being malicious. Below are descriptions of several strategies currently implemented to mitigate such attacks. Each strategy is associated with one or more parameters that can be adjusted by the user. Default settings for the parameters involved for each of these strategies have been chosen that are expected to work effectively for most users.
Basic Strategy #1: Decrease the Default User Influence
This strategy consists of decreasing or setting to zero the default score and / or confidence that is applied to all users. This has the same effect as if the seed user were to supply this rating to each and every profile that is known to exist. These parameters are located under the "User" tab and may be adjusted therein. You may wish to forego this strategy and set default user influence to a higher value if there are a large number of unvetted users who provide useful ratings information, with a relatively low percentage of bad actors.
Decrease the Attenuation Factor
ATTENUATION FACTOR: slider moved to basicGrapevineControlsTopPanel.php
The weight of each * user-to-user trust rating is attenuated by multiplying it by the Attenuation Factor, which is a parameter that may be adjusted continuously between 0 and 1. By decreasing the attenuation factor, the farther away a user is from you in the grapevine (number of connections), the less influence that user has. This has the effect of decreasing the "Size" of your grapevine.
* Ratings issued by the Seed User are not multiplied by the Attenuation Factor.
Special Strategy #1: Reference User
m1sf:
OFF
ON
All trust scores are (more appropriately) interpreted using the rater as the reference (rather than the poorly-defined "average user"). This practice is conceptually more appropriate (because the notion of "average user" is problematic). This effectively sharpens the filter between low-trust and high-trust groups. (However, in some ways it may increase vulnerability to attacks from bad actors?)
Special Strategy #2: Bidirectionality
m2sf:
OFF
ON
Trust ratings affect the rater's average as well as the ratee's. Conceptually this is more appropriate, and mitigates certain types of attacks (malicious user accounts upvoting each other within a group to bootstrap reputation).
Special Strategy #3: Reign in Sky-High Ratings
m3sf:
OFF
ON
Unreasonably high trust ratings are given less weight. This mitigates attacks certain types of attacks (malicious user accounts upvoting other malicious accounts).
m4sf:
OFF
ON
a parameter to modify the strength of Strategy #3.
m5sf:
strict
loose
a parameter to specify Strategy #3's boundary between a "reasonable" and an "unreasonablly high" credibility score.
OVERVIEW SETTINGS TAB
anchor
attenuation factor
default user settings
default Average Score for Users:
?
The default Average Score for Users: Between 0 and 1, this is how much weight you would attribute
to a user that doesn't have any ratings. Setting the default low is one way to stratify users: the fewer ratings there are, the closer the final score is to the default.
To turn this feature off, set the "Confidence" slider to zero.
Confidence: This indicates how much weight you want to throw behind the default Average Score.
Setting confidence to zero effectively turns off this feature. The higher the Confidence, the more actual ratings it will take to move the final score away from the default.
default User Influence:
X
?
When calculating weighted average scores, the rater's Influence is the primary determinant of the weight.
Influence is calculated as the user's Average Score multiplied by the Confidence in that score. (Sometimes with a few other modifications.)
User Influence ?
username
Vendor Trust ?
Moderator Trust ?
# of flags
flags
User Influence
username
Vendor Trust
Moderator Trust
# of flags
flags
Reset & Recalculate User Trust Scores
default Average Score for Listings:
?
The default Average Score for Listings: Between 0 and 5, this is the number of stars to attribute
to a listing that doesn't have any ratings. Setting the default low is one way to stratify listings: the more (higher) ratings come in, the higher the final score.
To turn this feature off, set the "Confidence" slider to zero.
Confidence: This indicates how much weight you want to throw behind the default Average Score.
Setting Confidence to zero gives no weight to the Average Score and effectively turns this feature off.
The higher the Confidence, the more actual ratings it will take to move the final score away from the default.
See what listings have been flagged by my Grapevine
listsOfAllListingFlags
Weighted Average (Grapevine)
Simple Average
Listing
Vendor
# of ratings
input (certainty)
# of flags
flags
Weighted Average (Grapevine)
Simple Average
Listing
Vendor
# of ratings
input (certainty)
# of flags
flags
Polls