Update of "Experiment: mmmv_linguistic_anonymizer_t1"
Not logged in

Artifact ID: fea1d5427082bf2bdbb27b57613b1bf37d336fa8
Page Name:Experiment: mmmv_linguistic_anonymizer_t1
Date: 2019-01-06 00:34:58
Original User: martin_vahi
Parent: 5e874d63888c2656c8de33657766499b4105bba8 (diff)

This experiment is currently so incomplete that it lacks both, the code and a thorough specification candidate.

The Problem

People tend to have relatively unique expression styles, which are influenced by the person's knowledge and life experience. For example, malware that uses some mathematics theory T_1 can not be written by a person, who does not know the theory T_1, unless the malware uses an interface of a component that is written by a person, who knows the theory T_1. The same with "general talk". There is no point of publishing through some "WikiLeaks", if the published text itself reveals the "leaker". 

Solution Idea

If the person, who writes the text, can be identified by the writing style, then a partial solution might be that the "leaker" (hereafter P_leaker) teaches an other person (hereafter: P_public) the material that the P_leaker wants to be leaked and the P_public writes the publishable text that contains the data that is subject to leaking. That solution does not offer anonymity to P_leaker even, if the P_public were skilled, willing and rigorous enough to keep the identity of the P_leaker a secret, because the P_public can be captured and set to delirium with chemicals, unless the P_public is not a real person, but just a copy of a software agent, a lot like speech synthesis software is. As meaning depends on context and the context is determined by the available data that describes the context, the agent instances of different leakers may start to differ, consequently allowing the identification of the P_leaker, but that issue can be mitigated by making "personality data", world_knowledge, context of the agent P_public available over some anonymization network, may be Tor, if there are enough users than may be even some plain P2P file sharing network. 

Many of the subcomponents of such an agent, the P_public, are probably the same as are the subcomponents of the linguistics based search engine. Therefore, by working on a modular version of the linguistics based search engine one works also on that linguistics based anonymization software.