Publications

The Director's Office at MIT Lincoln Laboratory (MIT LL) requested a comprehensive study on artificial intelligence (AI) focusing on present applications and future science and technology (S&T) opportunities in the Cyber Security and Information Sciences Division (Division 5). This report elaborates on the main results from the study. Since the AI field is evolving so rapidly, the study scope was to look at the recent past and ongoing developments to lead to a set of findings and recommendations. It was important to begin with a short AI history and a lay-of-the-land on representative developments across the Department of Defense (DoD), intelligence communities (IC), and Homeland Security. These areas are addressed in more detail within the report. A main deliverable from the study was to formulate an end-to-end AI canonical architecture that was suitable for a range of applications. The AI canonical architecture, formulated in the study, serves as the guiding framework for all the sections in this report. Even though the study primarily focused on cyber security and information sciences, the enabling technologies are broadly applicable to many other areas. Therefore, we dedicate a full section on enabling technologies in Section 3. The discussion on enabling technologies helps the reader clarify the distinction among AI, machine learning algorithms, and specific techniques to make an end-to-end AI system viable. In order to understand what is the lay-of-the-land in AI, study participants performed a fairly wide reach within MIT LL and external to the Laboratory (government, commercial companies, defense industrial base, peers, academia, and AI centers). In addition to the study participants (shown in the next section under acknowledgements), we also assembled an internal review team (IRT). The IRT was extremely helpful in providing feedback and in helping with the formulation of the study briefings, as we transitioned from datagathering mode to the study synthesis. The format followed throughout the study was to highlight relevant content that substantiates the study findings, and identify a set of recommendations. An important finding is the significant AI investment by the so-called "big 6" commercial companies. These major commercial companies are Google, Amazon, Facebook, Microsoft, Apple, and IBM. They dominate in the AI ecosystem research and development (R&D) investments within the U.S. According to a recent McKinsey Global Institute report, cumulative R&D investment in AI amounts to about $30 billion per year. This amount is substantially higher than the R&D investment within the DoD, IC, and Homeland Security. Therefore, the DoD will need to be very strategic about investing where needed, while at the same time leveraging the technologies already developed and available from a wide range of commercial applications. As we will discuss in Section 1 as part of the AI history, MIT LL has been instrumental in developing advanced AI capabilities. For example, MIT LL has a long history in the development of human language technologies (HLT) by successfully applying machine learning algorithms to difficult problems in speech recognition, machine translation, and speech understanding. Section 4 elaborates on prior applications of these technologies, as well as newer applications in the context of multi-modalities (e.g., speech, text, images, and video). An end-to-end AI system is very well suited to enhancing the capabilities of human language analysis. Section 5 discusses AI's nascent role in cyber security. There have been cases where AI has already provided important benefits. However, much more research is needed in both the application of AI to cyber security and the associated vulnerability to the so-called adversarial AI. Adversarial AI is an area very critical to the DoD, IC, and Homeland Security, where malicious adversaries can disrupt AI systems and make them untrusted in operational environments. This report concludes with specific recommendations by formulating the way forward for Division 5 and a discussion of S&T challenges and opportunities. The S&T challenges and opportunities are centered on the key elements of the AI canonical architecture to strengthen the AI capabilities across the DoD, IC, and Homeland Security in support of national security.

The processor of a drone runs essential functions of sensing, communications, coordination, and control. This is the conventional view. But in today's cyber environment, the processor must also provide security to assure mission completion. We have been developing a secure processing architecture for mission assurance. A study on state-of-the-art secure processing technologies has revealed that no one-size-fits-all solution can fully meet our requirements. In fact, we have concluded that the provision of a secure processor as a mission assurance foundation must be holistic and should be approached from a systems perspective. We have thus applied a systems analysis approach to create a secure base for the system. This paper describes our journey of adapting and synergizing various secure processing technologies into a baseline asymmetric multicore processing architecture. We will also describe a functional and security co-design environment, created to customize and optimize the architecture in a design space consisting of hardware, software, performance, and assurance.

Modern mobile devices place a wide variety of sensors and services within the personal space of their users. As a result, these devices are capable of transparently monitoring many sensitive aspects of these users' lives (e.g., location, health, or correspondences). Users typically trade access to this data for convenient applications and features, in many cases without a full appreciation of the nature and extent of the information that they are exposing to a variety of third parties. Nevertheless, studies show that users remain concerned about their privacy and vendors have similarly been increasing their utilization of privacy-preserving technologies in these devices. Still, despite significant efforts, these technologies continue to fail in fundamental ways, leaving users' private data exposed. In this work, we survey the numerous components of mobile devices, giving particular attention to those that collect, process, or protect users' private data. Whereas the individual components have been generally well studied and understood, examining the entire mobile device ecosystem provides significant insights into its overwhelming complexity. The numerous components of this complex ecosystem are frequently built and controlled by different parties with varying interests and incentives. Moreover, most of these parties are unknown to the typical user. The technologies that are employed to protect the users' privacy typically only do so within a small slice of this ecosystem, abstracting away the greater complexity of the system. Our analysis suggests that this abstracted complexity is the major cause of many privacy-related vulnerabilities, and that a fundamentally new, holistic, approach to privacy is needed going forward. We thus highlight various existing technology gaps and propose several promising research directions for addressing and reducing this complexity.

Protecting enterprise networks requires continuous risk assessment that automatically identifies and prioritizes cyber security risks, enables efficient allocation of cyber security resources, and enhances protection against modern cyber threats. Lincoln Laboratory created a network security model to guide the development of such risk assessments and, for the most important cyber threats, designed practical risk metrics that can be computed automatically and continuously from security-relevant network data.

We present PANDA, an open-source tool that has been purpose-built to support whole system reverse engineering. It is built upon the QEMU whole system emulator, and so analyses have access to all code executing in the guest and all data. PANDA adds the ability to record and replay executions, enabling iterative, deep, whole system analyses. Further, the replay log files are compact and shareable, allowing for repeatable experiments. A nine billion instruction boot of FreeBSD, e.g., is represented by only a few hundred MB. Furhter, PANDA leverages QEMU's support of thirteen different CPU architectures to make analyses of those diverse instruction sets possible within the LLVM IR. In this way, PANDA can have a single dynamic taint analysis, for example, that precisely supports many CPUs. PANDA analyses are written in a simple plugin architecture which includes a mechanism to share functionality between plugins, increasing analysis code re-use and simplifying complex analysis development. We demonstrate PANDA's effectiveness via a number of use cases, including enabling an old but legitimate version of Starcraft to rund espite a lost CD key, in-depth diagnosis of an Internet Explorer crash, and uncovering the censorship activities and mechanisms of a Chinese IM client.

This work presents the results of a three-year project that assessed nine different privacy-preserving data search systems. We detail the design of a software assessment framework that focuses on low system footprint, repeatability, and reusability. A unique achievement of this project was the automation and integration of the entire test process, from the production and execution of tests to the generation of human-readable evaluation reports. We synthesize our experiences into a set of simple mantras that we recommend following in the design of any assessment framework.

Big data volume continues to grow at unprecedented rates. One of the key features that makes big data valuable is the promise to find unknown patterns or correlations that may be able to improve the quality of processes or systems. Unfortunately, with the exponential growth in data, users often have difficulty in visualizing the often-unstructured, non-homogeneous data coming from a variety of sources. The recent growth in popularity of 3D printing has ushered in a revolutionary way to interact with big data. Using a 3D printed mockup up a physical or notional environment, one can display data on the mockup to show real-time data patterns. In this poster and demonstration, we describe the process of 3D printing and demonstrate an application of displaying Twitter data on a 3D mockup of the Massachusetts Institute of Technology (MIT) campus, known as LuminoCity.

In this work, we describe LuminoCity, a novel three-dimensional data display. A 3D printed model of Cambridge, MA was generated from LADAR data. A translucent plastic model was then cast from a mold of the 3D printed model. We developed a display system to project data onto the translucent model, and we can project a wide range of analyses onto the city, including satellite imagery and network traffic.

Dynamic information flow tracking is a well-known dynamic software analysis technique with a wide variety of applications that range from making systems more secure, to helping developers and analysts better understand the code that systems are executing. Traditionally, the fine-grained analysis capabilities that are desired for the class of these systems which operate at the binary level require tight coupling to a specific ISA. This places a heavy burden on developers of these systems since significant domain knowledge is required to support each ISA, and the ability to amortize the effort expended on one ISA implementation cannot be leveraged to support other ISAs. Further, the correctness of the system must carefully evaluated for each new ISA. In this paper, we present a general approach to information flow tracking that allows us to support multiple ISAs without mastering the intricate details of each ISA we support, and without extensive verification. Our approach leverages binary translation to an intermediate representation where we have developed detailed, architecture-neutral information flow models. To support advanced instructions that are typically implemented in C code in binary translators, we also present a combined static/dynamic analysis that allows us to accurately and automatically support these instructions. We demonstrate the utility of our system in three different application settings: enforcing information flow policies, classifying algorithms by information flow properties, and characterizing types of programs which may exhibit excessive information flow in an information flow tracking system.

The Micro-sized Microwave Atmospheric Satellite (MicroMAS) is a 3U cubesat (34x10x10 cm, 4.5 kg) hosting a passive microwave spectrometer operating near the 118.75-GHz oxygen absorption line. The focus of the first MicroMAS mission (hereafter, MicroMAS-1) is to observe convective thunderstorms, tropical cyclones, and hurricanes from a near-equatorial orbit at approximately 500-km altitude. A MicroMAS flight unit is currently being developed in anticipation of a 2014 launch. A parabolic reflector is mechanically rotated as the spacecraft orbits the earth, thus directing a cross-track scanned beam with FWHM beamwidth of 2.4-degrees, yielding an approximately 20-km diameter footprint at nadir incidence from a nominal altitude of 500 km. Radiometric calibration is carried out using observations of cold space, the earth?s limb, and an internal noise diode that is weakly coupled through the RF front-end electronics. A key technology feature is the development of an ultra-compact intermediate frequency processor module for channelization, detection, and A-to-D conversion. The antenna system and RF front-end electronics are highly integrated and miniaturized. A MicroMAS-2 mission is currently being planned using a multiband spectrometer operating near 118 and 183 GHz in a sunsynchronous orbit of approximately 800-km altitude. A HyMAS- 1 (Hyperspectral Microwave Atmospheric Satellite) mission with approximately 50 channels near 118 and 183 GHz is also being planned. In this paper, the mission concept of operations will be discussed, the radiometer payload will be described, and the spacecraft subsystems (avionics, power, communications, attitude determination and control, and mechanical structures) will be summarized.