Resumen de Information leakage and steganography: detecting and blocking covert channels
This PhD Thesis explores the threat of information theft perpetrated by malicious insiders. As opposite to outsiders, insiders have access to information assets belonging the organization, know the organization infrastructure and more importantly, know the value of the different assets the organization holds. The risk created by malicious insiders have led both the research community and commercial providers to spend efforts on creating mechanisms and solutions to reduce it. However, the lack of certain controls by current proposals may led security administrators to a false sense of security that could actually ease information theft attempts.
As a first step of this dissertation, a study of current state of the art proposals regarding information leakage protections has been performed. This study has allowed to identify the main weaknesses of current proposals which are mainly the usage of steganographic algorithms, the lack of control of modern mobile devices and the lack of control of the action the insiders perform inside the different trusted applications they commonly use. Each of these drawbacks have been explored during this dissertation.
Regarding the usage of steganographic algorithms, two different steganographic systems have been proposed. First, a steganographic algorithm that transforms source code into innocuous text has been presented. This system uses free context grammars and to parse the source code to be hidden and produce an innocuous text. This system could be used to extract valuable source code from software development environments, where security restrictions are usually softened. Second, a steganographic application for iOS devices has also been presented. This application, called ``Hide It In'' allows to embed images into other innocuous images and send those images through the device email account. This application includes a cover mode that allows to take pictures without showing that fact in the device screen. The usage of these kinds of applications is suitable in most of the environments which handle sensitive information, as most of them do not incorporate mechanisms to control the usage of advanced mobile devices. The application, which is already available at the Apple App Store, has been downloaded more than 5.000 times.
In order to protect organizations against the malicious usage of steganography, several techniques can be implemented. In this thesis two different approaches are presented. First, steganographic detectors could be deployed along the organization to detect possible transmissions of stego-objects outside the organization perimeter. In this regard, a proposal to detect hidden information inside executable files has been presented. The proposed detector, which measures the assembler instruction selection made by compilers, is able to correctly identify stego-objects created through the tool \emph{Hydan}. Second, steganographic sanitizers could be deployed over the organization infrastructure to reduce the capacity of covert channels that can transmit information outside the organization. In this regard, a framework to avoid the usage of steganography over the HTTP protocol has been proposed. The presented framework, diassembles HTTP messages, overwrites the possible carriers of hidden information with random noise and assembles the HTTP message again. Obtained results show that it is possible to highly reduce the capacity of covert channels created through HTTP. However, the system introduces a considerable delay in communications.
Besides steganography, this thesis has also addressed the usage of trusted applications to extract information from organizations. Although applications execution inside an organization can be restricted, trusted applications used to perform daily tasks are generally executed without any restrictions. However, the complexity of such applications can be used by an insider to transform information in such a way that deployed information protection solutions are not able to detect the transformed information as sensitive. In this thesis, a method to encrypt sensitive information using trusted applications is presented. Once the information has been encrypted it is possible to extract it outside the organization without raising any alarm in the deployed security systems. This technique has been successfully evaluated against a state of the art commercial data leakage protection solution. Besides the presented evasion technique, several improvements to enhance the security of current DLP solutions are presented. These are specifically focused in avoiding information leakage through the usage of trusted applications.
The contributions of this dissertation have shown that current information leakage protection mechanisms do not fully address all the possible attacks that a malicious insider can commit to steal sensitive information. However, it has been shown that it is possible to implement mechanisms to avoid the extraction of sensitive information by malicious insiders. Obviously, avoiding such attacks does not mean that all possible threats created by malicious insiders are addressed. It is necessary then, to continue studying the threats that malicious insiders pose to the confidentiality of information assets and the possible mechanisms to mitigate them.
