The State of Reputation: Part 2


(adsbygoogle = window.adsbygoogle || []).push({});

(function ($) {
var bsaProContainer = $(‘.bsaProContainer-2’);
var number_show_ads = “0”;
var number_hide_ads = “0”;
if ( number_show_ads > 0 ) {
setTimeout(function () { bsaProContainer.fadeIn(); }, number_show_ads * 1000);
}
if ( number_hide_ads > 0 ) {
setTimeout(function () { bsaProContainer.fadeOut(); }, number_hide_ads * 1000);
}
})(jQuery);

In the previous part of this three-part blog series, we detailed the current state of Digital Reputation, and the tactics used that undermine a significant proportion of the market, leaving ordinary people confused as to what to believe and what not to believe. In this part, we will cover how the market currently tries to manage the problem, and why it is failing.

.bsaProContainer-4 {
display: none;
}

(function ($) {
var bsaProContainer = $(‘.bsaProContainer-4’);
var number_show_ads = “3”;
var number_hide_ads = “0”;
if ( number_show_ads > 0 ) {
setTimeout(function () { bsaProContainer.fadeIn(); }, number_show_ads * 1000);
}
if ( number_hide_ads > 0 ) {
setTimeout(function () { bsaProContainer.fadeOut(); }, number_hide_ads * 1000);
}
})(jQuery);

As detailed in the previous part of this series, there are seven common ways fake or false feedback & reviews are created:

  • ‘Crowdturfing’
  • Paid for Reviews
  • Bots
  • ‘Sock Puppeting’
  • ‘Blackhatting’ Competitors
  • Social Engineering
  • Punishment Reviews

Each of these is dealt with using a mixture of techniques, and they are woefully inadequate for the task facing them. Despite a lot of marketing spin, technical double talk and reassuring noises, the tools used to police these systems are akin to trying to put out a fire with a water pistol.

(adsbygoogle = window.adsbygoogle || []).push({});

(function ($) {
var bsaProContainer = $(‘.bsaProContainer-‘);
var number_show_ads = “0”;
var number_hide_ads = “0”;
if ( number_show_ads > 0 ) {
setTimeout(function () { bsaProContainer.fadeIn(); }, number_show_ads * 1000);
}
if ( number_hide_ads > 0 ) {
setTimeout(function () { bsaProContainer.fadeOut(); }, number_hide_ads * 1000);
}
})(jQuery);

So, to pull back the curtain, these are the techniques currently in use today:

Blacklisting

How It Is Done: Currently, blacklisting is the most common approach. It involves trying to flag IP addresses and accounts for commonality of action- for example, watching 300 accounts ‘brigade’ (Act as a coordinated swarm) to submit fake reviews, or seeing an excessive number of reviews from an IP address. Some also throw in browser fingerprinting (in a slightly ‘invade your privacy/quite creepy’ way) on top of this.

(adsbygoogle = window.adsbygoogle || []).push({});

(function ($) {
var bsaProContainer = $(‘.bsaProContainer-1’);
var number_show_ads = “0”;
var number_hide_ads = “0”;
if ( number_show_ads > 0 ) {
setTimeout(function () { bsaProContainer.fadeIn(); }, number_show_ads * 1000);
}
if ( number_hide_ads > 0 ) {
setTimeout(function () { bsaProContainer.fadeOut(); }, number_hide_ads * 1000);
}
})(jQuery);

Where It Breaks: Blacklisting is a ‘band-aid for a bullet wound’- it doesn’t stop fake reviews, it slows the flow of fake reviews from sources using crude network-level filtering, and undermines security for those using VPNs or Proxies to ensure their privacy. Compounding that is that is very easy to evade by those who are intent on making money from fake reviews- so it doesn’t solve the problem, makes life harder for users and engages in dubious techniques.

Manual Review

How It Is Done: Some unfortunate soul has to vet the reviews manually, checking for commonality, inconsistencies, and trying to find fakes hidden amongst the real reviews.

(adsbygoogle = window.adsbygoogle || []).push({});

(function ($) {
var bsaProContainer = $(‘.bsaProContainer-1’);
var number_show_ads = “0”;
var number_hide_ads = “0”;
if ( number_show_ads > 0 ) {
setTimeout(function () { bsaProContainer.fadeIn(); }, number_show_ads * 1000);
}
if ( number_hide_ads > 0 ) {
setTimeout(function () { bsaProContainer.fadeOut(); }, number_hide_ads * 1000);
}
})(jQuery);

Where It Breaks: Manual review is resource intensive- humans aren’t cheap, and even then the personnel hired at the lower end of the pay bracket- they are not experts in linguistics trained at Quantico, they are large teams working long hours with no thanks. Worse, not all reviews are subjected to the same scrutiny- to save of the resource costs, only when items are flagged are they subjected to manual review. When was the last time you flagged a review you thought was suspicious? It’s likely that less than 1% are flagged, and when the problem accounts for 20% of reviews — that leaves a huge disparity.

Natural Language Processing & AI

How It Is Done: AI & NLP have been held up as the answer to Fake Reviews for going on five years now. The technique typically uses Natural Language Processing algorithms (and sometimes ‘algorithm’ is overstating it) to analyse the review as written, as well as using Machine Learning in combination with similar techniques as Blacklisting to arrive at an automated solution that screens reviews both at submission and afterwards.

Where It Breaks: The problem with this approach is it depends on how good your AI is. Unlike Hollywood ideas of AI, AI is still in the very early (and quite stupid) stage of development. Worse, the fakers use the same techniques- and contrary to the fictional idea of two AIs fighting it out like Godzilla vs Mothra, it’s an uneven contest. The defending AI has to cover any and all possible attack vectors- the attacking AI only needs to find one hole, and it can undermine it all.

In August, the number of reviews on Amazon that, after manual review, were flagged as unnatural more than doubled- in other words, the faker AI techniques are winning and are getting passed AI developed by a company with virtually limitless resources who built their own AI platform for others to use.

If Amazon’s best minds cannot come up with an NLP/AI solution for the current review paradigm that is 100% effective, none of the smaller players can (Looking at you Yelp, TripAdvisor, E-Bay, etc). Worse, it’s less than 50% effective at the time of writing and shows no signs of improving.

Legal Action

How It Is Done: Find reviewers, serve a Cease & Desist and sue them individually. No. Really. Amazon has been doing this to reviewers offering fake reviews on Fiverr.

Where It Breaks: If you have to sue someone, your technology is broken. Ask the Music industry how litigation against Napster solved their piracy problem (Spoilers: It made it worse). It is a game of whack-a-mole targeting individuals who may or may not even be in a jurisdiction where you can stop them. It is the desperate act of a company that is tacitly acknowledging it cannot solve the problem.

Fighting the Flood

So, none of the techniques used address the problem adequate. Individual or in combination, none of these techniques deal with any of the common attacks, let alone when those attacks are used in combination (hint: they generally are).

So how do you fight this rising tide of synthetic feedback? That’s what we’ve spent nearly three years researching, designing and building- and we’ve got the answer. In the next part of this series we will show you how the Tru Reputation Network solves these problems and through doing so, opens up a bright future of possibilities that will affect every single user of the internet.


The State of Reputation: Part 2 was originally published in Tru Ltd on Medium, where people are continuing the conversation by highlighting and responding to this story.

Source link


(adsbygoogle = window.adsbygoogle || []).push({});

(function ($) {
var bsaProContainer = $(‘.bsaProContainer-2’);
var number_show_ads = “0”;
var number_hide_ads = “0”;
if ( number_show_ads > 0 ) {
setTimeout(function () { bsaProContainer.fadeIn(); }, number_show_ads * 1000);
}
if ( number_hide_ads > 0 ) {
setTimeout(function () { bsaProContainer.fadeOut(); }, number_hide_ads * 1000);
}
})(jQuery);

The post The State of Reputation: Part 2 appeared first on Cryptocurrency decentral market headlines.

Source: New feed

Click here for reuse options!
Copyright 2018 bitcoin4allfree.com

You may also like...

www.000webhost.com