Abstract #1
As the Internet of Things (IoT) becomes more prevalent in society, security threats continue to arise. Hackers are able to use many different types of attacks on IoT devices which can compromise either their functionality, or the privacy of their owners.
The first step to preventing these attacks is being able to identify the types of attacks, and the intentions behind them. Thus, the purpose of this research is two folds: first, to extend the Hidden Markov Model (HMM) to 2D-HMM and hierarchical-HMM in order to detect types of attacks and the intentions behind them. The second is to effectively apply the extended-HMM to recent threat identification and prediction in IoT.
Abstract #2
Internet of Things (IoT) attacks have rapidly risen in frequency in recent years as IoT devices become more commonplace in industry, businesses, and homes. Since these devices have very basic functionality and are not designed with security in mind, they are easy targets for attacks that can steal data or gain access to the network the devices are connected to.
Here we propose a tiered system of Hidden Markov Models (HMMs) for identifying these attacks and classifying them by type of attack. This system has a tree-based structure, with the main HMM being applied to the raw network data to identify attacks. This main HMM branches off into separate HMMs for each type of attack to classify the attacks according to how important the consequences of the attack are and how likely each attack is to happen.
Problem Statement
The Internet of Things is one of the largest advancements in technology in recent years, and is expected to become massively large in the near future. However, so many connected devices presents a large security concern: the internet is not secure and users with malicious intentions are able to compromise the integrity of IoT systems. Our primary goal is to develop multiple Hidden Markov Models to identify IoT attacks and the intentions behind them. We plan to apply this model to predict IoT attacks before they occur.
Intended Contributions
1.) 2D-HMM and Testing
A 2-dimensional HMM will use data collected from the network over time to predict the probability of different types of threats and the intent behind them.
3.) Classifying and identifying intent of threats in IoT from data
We will be using a 2-dimensional HMM to first identify the types of attacks on IoT, then to identify the intent behind the attacks (whether it is an attack on the network, control, controller, or data).
2.) Tiered-HMM and Verification
A tiered system of Hidden Markov Models will be implemented to detect different attacks and subattacks with greater accuracy than with a single HMM.
4.) Examining performance in terms of accuracy, precision, and computational efficiency
Applications of these 2-dimensional and tiered HMMs are only useful if they are not too intensive to run, and are able to correctly classify threats, so testing is required.