Machine Learning based Information Forensics from Smart Sources

Publication Type:
Issue Date:
Full metadata record
We live in a world that connects, socializes and interacts using internet. Humans generate tons of information on daily basis, according to Forbes 2.5 quintillion bytes of data is created each day in year 2018. The data creation pace is continuously accelerating with the growth of the Internet of Things (IoT). Extensive social media usage fuels data creation which is primarily generated from mobile phones. In the present scenario, data is the asset and this asset is extremely vulnerable. In our research work we utilized these data sources to aid digital forensics investigation. Due to our technology engulfed lifestyle, we leave a lot of information about ourselves during our routine activities. These traces are used by the wrongdoers for their vicious objectives but also these traces can be used by investigators to understand any incident and to penalize the delinquents. Forensics investigators face challenges with a huge amount of data during investigations. Whether the data source is an online social network, a smart phone or an IOT based environment, huge amount of data adds complexity and delay to forensics investigation. To contribute to the forensics investigation we propose the use of machine learning for forensics data analysis. Forensics investigation is a three-phase process including data acquisition, data analysis and presentation. Our research focuses on the first two phases of the forensics investigation cycle i.e. data collection and data analysis. This thesis discusses following research achievements: 1. Data acquisition from smart sources for forensic information especially for IoT 2. Machine learning based data analysis to extract forensic artefacts 3. IoT forensics framework (acquisition and analysis phase) implementation This thesis is segmented based on the three data sources for data analysis namely online social networks, smart phones and sensor-based networks (IoT). Using IoT based data this thesis proposes a scheme SACIFS (Smart aged care information forensics) for IoT forensic feature extraction and data analysis of elderly patients monitored in a nursing home environment. We developed our machine learning model based on Support Vector Machine to detect an incident and highlight relevant forensic artefacts. We used smart phone data in Digital Forensic Intelligence Analysis Cycle framework to identify strongly connected contacts during triage phase. We classified the contacts of a smart phone user with respect to their closeness, extracted from the data features from Facebook messenger. Moreover, utilizing publically available Online Social Network, we analysed multiple tools to collect, analyse and visualize data from Facebook pages and groups.
Please use this identifier to cite or link to this item: