Aspects of internet security: identity management and online child protection

This thesis examines four main subjects; consumer federated Internet Identity Management (IdM), text analysis to detect grooming in Internet chat, a system for using steganographed emoticons as ‘digital fingerprints’ in instant messaging and a systems analysis of online child protection. The Internet was never designed to support an identity framework. The current username / password model does not scale well and with an ever increasing number of sites and services users are suffering from password fatigue and using insecure practises such as using the same password across websites. In addition users are supplying personal information to vast number of sites and services with little, if any control over how that information is used. A new identity metasystem promises to bring federated identity, which has found success in the enterprise to the consumer, placing the user in control and limiting the disclosure of personal information. This thesis argues though technical feasible no business model exists to support consumer IdM and without a major change in Internet culture such as a breakdown in trust and security a new identity metasystem will not be realised. Is it possible to detect grooming or potential grooming from a statistical examination of Internet chat messages? Using techniques from speaker verification can grooming relationships be detected? Can this approach improve on the leading text analysis technique – Bayesian trigram analysis? Using a novel feature extraction technique and Gaussian Mixture Models (GMM) to detect potential grooming proved to be unreliable. Even with the benefit of extensive tuning the author doubts the technique would match or improve upon Bayesian analysis. Around 80% of child grooming is blatant with the groomer disguising neither their age nor sexual intent. Experiments conducted with Bayesian trigram analysis suggest this could be reliably detected, detecting the subtle, devious remaining 20% is considerably harder and reliable detection is questionable especially in systems using teenagers (the most at risk group). Observations of the MSN Messenger service and protocol lead the author to discover a method by which to leave digitally verifiable files on the computer of anyone who chats with a child by exploiting the custom emoticon feature. By employing techniques from steganography these custom emoticons can be made to appear innocuous. Finding and removing custom emoticons is a non-trivial matter and they cannot be easily spoofed. Identification is performed by examining the emoticon (file) hashes. If an emoticon is recovered e.g. in the course of an investigation it can be hashed and the hashed compared against a database of registered users and used to support non-repudiation and confirm if an individual has indeed been chatting with a child. Online child protection has been described as a classic systems problem. It covers a broad range of complex, and sometimes difficult to research issues including technology, sociology, psychology and law, and affects directly or indirectly the majority of the UK population. Yet despite this the problem and the challenges are poorly understood, thanks in no small part to mawkish attitudes and alarmist media coverage. Here the problem is examined holistically; how children use technology, what the risks are, and how they can best be protected – based not on idealism, but on the known behaviours of children. The overall protection message is often confused and unrealistic, leaving parents and children ill prepared to protect themselves. Technology does have a place in protecting children, but this is secondary to a strong and understanding parent/child relationship and education, both of the child and parent.