Recently we have seen the emergence new challenges to cyber security in which adversaries seek, not to simply to deny service, steal data, or utilize resources, but rather to use the network to distort information to influence opinion, thought, or action. Such Internet information manipulation is challenging in that the scale of the Internet makes it easier to influence large numbers of users, that the anonymity on the Internet makes attribution and identity hard, and that the control of information flow is distributed and highly automated allowing for easy modification of personalization and filtering algorithms.
This research project proposes to study information manipulation on the Internet. Specifically, it seeks to provide evidence of manipulation and demonstrate the serious nature of the problem. The project has three parts: (1) Constructing a corpus of documented cases of Internet information manipulation, (2) An identification of various players in Internet information flow and their relationships to each other, and (3) An empirical analysis of the scope of the information flow between these players. More broadly, the work will provide data to inform techniques, and methods that with enhance a truly interdisciplinary set of fields, from the social and political sciences, to networking and distributed systems, security, information retrieval, and data mining.
The Internet has been attributed with a staggering array of economic, social, and individual benefits. Recognizing that the smooth operation of the Internet is critical to continuing these benefits, researchers have spent considerable time and effort addressing rampant attacks such as identity theft, scams and fraud, and malware. Much of this work has focused on attacks that directly impact the dependability of the Internet as a distributed system (e.g., Availability, Reliability, Safety, Timeliness) or those attacks that threaten the confidentiality and integrity of the data the Internet contains. Recently, however, we have seen the emergence new challenges to cyber security in which adversaries seek, not to simply to deny service, steal data, or utilize resources, but rather to use the network to distort information to influence opinion, thought, or action. While such information manipulation has been discussed in political and social science contexts, the emergence of such deliberate activity on the Internet constitutes a real and present danger, which if left unchecked can harm our economy, culture, and democracy. The issues of scale, attribution, and distributed control provide unique challenges for researchers attempting to address this new form of information manipulation. This work sought to provide a quantitative and empirical grounding for the field of Internet information manipulation. As a launching point, we studied the prevalence of online tracking. Knowing how well third party networks can profile and use tracked data is an important consideration in the ongoing public debate on user privacy. We examined how web-browsing history can be used by ad networks in serving behavioral ads. Using our approach, we evaluated the effectiveness of online privacy tools that were designed to mitigate tracking and behavioral advertising. We examined a subset of tools that includes opt-out features, Do Not Track (DNT) headers and third party cookie blocking. We find that although both opt-out features and third party cookie blocking are largely effective at limiting behavioral advertising, DNT is not yet effective. Our work in measuring online information manipulation is influenced by a long running research agenda that seeks to measure the security of data online and noticeably impact our online security infrastructure. Support from this grant has been used to further this research in several notable ways. First, this grant has supported efforts in building new tools that enable fast and accurate mapping of services on the public IPv4 Internet. Our tool, ZMAP, is capable of providing accurate surveys of the full Internet in minutes. Second, data from this tool has been invaluable in studying a variety of important topics from measuring the ecosystem underlying confidential communication (e.g., TLS) on the Internet to studying emerging threats (e.g., Heartbleed).
