In the event of a cybercrime, it is necessary to examine the suspect’s digital device(s) in a forensic fashion so that the culprit can be presented in court along with the extracted evidence(s). But, factors such as existence and availability of anti-forensic tools/techniques and increasing replacement of hard disk drives with solid state disks have the ability to eradicate critical evidences and/or ruin their integrity. Therefore, having an alternative source of evidence with a lesser chance of being tampered with can be beneficial for the investigation. The organisational network traffic can fit into this role as it is an independent source of evidence and will contain a copy of all online user activities. Limitations of prevailing network traffic analysis techniques – packet based and flow based – are reflected as certain challenges in the investigation. The enormous volume and increasing encrypted nature of traffic, the dynamic nature of IP addresses of users’ devices, and the difficulty in extracting meaningful information from raw traffic are among those challenges. Furthermore, current network forensic tools, unlike the sophisticated computer forensic tools, are limited in their capability to exhibit functionalities such as collaborative working, visualisation, reporting and extracting meaningful user-level information. These factors increase the complexity of the analysis, and the time and effort required from the investigator.
The research goal was set to design a system that can assist in the investigation by minimising the effects of the aforementioned challenges, thereby reducing the cognitive load on the investigator, which, the researcher thinks, can take the investigator one step closer to the culprit. The novelty of this system comes from a newly proposed interaction based analysis approach, which will extract online user activities from raw network metadata. Practicality of the novel interaction-based approach was tested by designing an experimental methodology, which involved an initial phase of the researcher looking to identify unique signatures for activities performed on popular Internet applications (BBC, Dropbox, Facebook, Hotmail, Google Docs, Google Search, Skype, Twitter, Wikipedia, and YouTube) from the researcher’s own network metadata. With signatures obtained, the project moved towards the second phase of the experiment in which a much larger dataset (network traffic collected from 27 users for over 2 months) was analysed. Results showed that it is possible to extract unique signature of online user activities from raw network metadata. However, due to the complexities of the applications, signatures were not found for some activities. The interaction-based approach was able to reduce the data volume by eliminating the noise (machine to machine communication packets) and to find a way around the encryption issue by using only the network metadata.
A set of system requirements were generated, based on which a web based, client-server architecture for the proposed system (i.e. the User-Oriented Network Forensic Analysis Tool) was designed. The system functions in a case management premise while minimising the challenges that were identified earlier. The system architecture led to the development of a functional prototype. An evaluation of the system by academic experts from the field acted as a feedback mechanism. While the evaluators were satisfied with the system’s capability to assist in the investigation and meet the requirements, drawbacks such as inability to analyse real-time traffic and meeting the HCI standards were pointed out. The future work of the project will involve automated signature extraction, real-time processing and facilitation of integrated visualisation.
Date of Award | 2021 |
---|
Original language | English |
---|
Awarding Institution | |
---|
Supervisor | Nathan Clarke (Other Supervisor) |
---|
- Digital Forensics
- Network Forensics
- Insider Threat
- Online User Interactions
- Network Metadata Analysis
- Timeline of User Interactions
A Novel User Oriented Network Forensic Analysis Tool
Joy, D. (Author). 2021
Student thesis: PhD