Cody M. Lewis


A Brief Introduction to Modern Trust Models

Trust

Trust is a vital factor to establish before implementing any security system, it allows for detection of malware, faulty systems, malicious devices, among other features of a distributed system. It is subject to the situation it is applied, for example, you could trust most mechanics to repair a car, but would probably not trust them to perform surgery. The problem of trust establishment is equivalent to that of recommendation systems, therefore, any solution to either can be applied to the other. Although, there are many different approaches to solutions to each and they are subjective to the situation they are applied to.

Context

Modern trust models tend to consider context to a much greater extent, this is since in many modern applications of them, they are applied in big data situations. This means in order to predict which device within the system is the most trusted, they will learn and distinguish the context of what is is doing and what it has done. Context is dependent on the situation, but refers to the extra data outside of the rating of whether something is trusted, for example, in an Internet of Things network context may be the battery level of the thing and its computing power.

Observations and Records

When something happens within the system, the individuals contained will send a record of what has happened to all the others it wants to. These records will allow an individual to indirectly know what is happening within the system, giving them a broader view of the current events. An individual will then scale the records based on how much is trusts the sender, and will refer to its collection of both direct and indirect records in order to determine which individuals to trust.

Data Mining and Machine Learning

Once an individual has established a collection of records, it will then perform data mining techniques on them in order to filter which records are currently relevant, which ones to trust, and how much each of the other individuals are to be trusted. It will perform machine learning techniques to predict which providers are to be the most trusted to perform the requested service.


RSS ⬆ Back to top