Publications

Monitoring and Managing High Performance Computing (HPC) systems and environments generate an ever growing amount of data. Making sense of this data and generating a platform where the data can be visualized for system administrators and management to proactively identify system failures or understand the state of the system requires the platform to be as efficient and scalable as the underlying database tools used to store and analyze the data. In this paper we will show how we leverage Accumulo, d4m, and Unity to generate a 3D visualization platform to monitor and manage the Lincoln Laboratory Supercomputer systems and how we have had to retool our approach to scale with our systems.

We are on the cusp of a computational renaissance in space, and we should not bring past terrestrial missteps along. Commercial off-the-shelf (COTS) processors -- much more powerful than traditional rad-hard devices -- are increasingly used in a variety of low-altitude, short-duration CubeSat class missions. With this new-found headroom, the incessant drumbeat of "faster, cheaper, faster, cheaper" leads a familiar march towards Linux and a menagerie of existing software packages, each more bloated and challenging to secure than the last. Lincoln Laboratory has started a pilot effort to design and prototype an exemplar secure satellite processing platform, initially geared toward CubeSats but with a clear path to larger missions and future high performance rad-hard processors. The goal is to provide engineers a secure "grab-and-go" architecture that doesn't unduly hamstring aggressive build timelines yet still provides a foundation of security that can serve adopting systems well, as well as future systems derived from them. This document lays out the problem space for cybersecurity in this domain, derives design guidelines for future secure space systems, proposes an exemplar architecture that implements the guidelines, and provides a solid starting point for near-term and future satellite processing.

Analyzing large scale networks requires high performance streaming updates of graph representations of these data. Associative arrays are mathematical objects combining properties of spreadsheets, databases, matrices, and graphs, and are well-suited for representing and analyzing streaming network data. The Dynamic Distributed Dimensional Data Model (D4M) library implements associative arrays in a variety of languages (Python, Julia, and Matlab/Octave) and provides a lightweight in-memory database. Associative arrays are designed for block updates. Streaming updates to a large associative array requires a hierarchical implementation to optimize the performance of the memory hierarchy. Running 34,000 instances of a hierarchical D4M associative arrays on 1,100 server nodes on the MIT SuperCloud achieved a sustained update rate of 1,900,000,000 updates per second. This capability allows the MIT SuperCloud to analyze extremely large streaming network data sets.

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.

Accepting input from the outside world is one of the most dangerous things a system can do. Since type information is lost across system boundaries, systems must perform type-specific input handling routines to recover this information. Adversaries can carefully craft input data to exploit any bugs or vulnerabilities in these routines, thereby causing dangerous memory errors. Including input validation routines in kernels is especially risky. Sensitive memory contents and powerful privileges make kernels a preferred target of attackers. Furthermore, the fact that kernels must process user input, network data, as well as input from a wide array of peripheral devices means that including such input validation schemes is unavoidable. In this thesis we present Automatic Validation of Input Data (AVID), which helps solve the issue of input validation within kernels by automatically generating parser implementations for developer-defined structs. AVID leverages not only the unambiguity guarantees of parsing expression grammars but also the type safety guarantees of Rust. We show how AVID can be used to resolve a manufactured vulnerability in Tock, an operating system written in Rust for embedded systems. Using Rust’s procedural macro system, AVID generates parser implementations at compile time based on existing Rust struct definitions. AVID exposes a simple and convenient parser API that is able to validate input and then instantiate structs from the validated input. AVID's simple interface makes it easy for developers to use and to integrate with existing codebases.

Computer programmers use custom memory allocators as an alternative to built-in or general-purpose memory allocators with the intent to improve performance and minimize human error. However, it is difficult to achieve both memory safety and performance gains on custom memory allocators. In this thesis, we study the relationship between memory safety and custom allocators. We analyze three popular servers, Apache, Nginx, and Appweb, and show that while the performance benefits might exist in the unprotected version of the server, as soon as partial or full memory safety is enforced, the picture becomes much more complex. Based on the target, using a custom memory allocator might be faster, about the same, or slower than the system memory allocator. Another caveat is that custom memory allocation can only be protected partially (at the allocation granularity) without manual modification. In addition, custom memory allocators may also introduce additional vulnerabilities to an application (e.g., OpenSSL Heartbleed). We thus conclude that using custom memory allocators is very nuanced, and that the challenges they pose may outweigh the small performance gains in the unprotected mode in many cases. Our findings suggest that developers must carefully consider the trade-offs and caveats of using a custom memory allocator before deploying it in their project.

This chapter continues the topic we started to discuss in the previous chapter – the human factors. However, it focuses on a specific method of enhancing cyber resilience via establishing appropriate rules for employees of an organization under consideration. Such rules aim at reducing threats from, for example, current or former employees, contractors, and business partners who intentionally use their authorized access to an organization to harm the organization. System users can also unintentionally contribute to cyber-attacks, or themselves become a passive target of a cyber-attack. The implementation of work-related rules is intended to decrease such risks. However, rules implementation can also increase the risks that arise from employee disregard for rules. This can occur when the rules become too restrictive, and employees become more likely to disregard the rules. Furthermore, the more often employees disregard the rules both intentionally and unintentionally, the more likely insider threats are able to observe and mimic employee behavior. This chapter shows how to find an intermediate, optimal collection of rules between the two extremes of "too many rules" and "not enough rules."

Black box machine learning methods have allowed researchers to design accurate models using large amounts of data at the cost of interpretability. Model interpretability not only improves user buy-in, but in many cases provides users with important information. Especially in the case of the classification problems addressed in this thesis, the ideal model should not only provide accurate predictions, but should also inform users of how features affect the results. My research goal is to solve real-world problems and compare how different classification models affect the outcomes and interpretability. To this end, this thesis is divided into two parts: food safety risk analysis and human trafficking detection. The first half analyzes the characteristics of supermarket suppliers in China that indicate a high risk of food safety violations. Contrary to expectations, supply chain dispersion, internal inspections, and quality certification systems are not found to be predictive of food safety risk in our data. The second half focuses on identifying human trafficking, specifically sex trafficking, advertisements hidden amongst online classified escort service advertisements. We propose a novel but interpretable keyword detection and modeling pipeline that is more accurate and actionable than current neural network approaches. The algorithms and applications presented in this thesis succeed in providing users with not just classifications but also the characteristics that indicate food safety risk and human trafficking ads.

Human trafficking is a form of modern-day slavery affecting an estimated 40 million victims worldwide, primarily through the commercial sexual exploitation of women and children. In the last decade, the advertising of victims has moved from the streets to websites on the Internet, providing greater efficiency and anonymity for sex traffickers. This shift has allowed traffickers to list their victims in multiple geographic areas simultaneously, while also improving operational security by using multiple methods of electronic communication with buyers; complicating the ability of law enforcement to disrupt these illicit organizations. In this paper, we address this issue and present a novel unsupervised and scalable template matching algorithm for analyzing and detecting complex organizations operating on adult service websites. The algorithm uses only the advertisement content to uncover signature patterns in text that are indicative of organized activities and organizational structure. We apply this method to a large corpus of adult service advertisements retrieved from backpage.com, and show that the networks identified through the algorithm match well with surrogate truth data derived from phone number networks in the same corpus. Further exploration of the results show that the proposed method provides deeper insights into the complex structures of sex trafficking organizations, not possible through networks derived from phone numbers alone. This method provides a powerful new capability for law enforcement to more completely identify and gather evidence about trafficking networks and their operations.

This paper explains the process by which Intel Software Guard Extensions (SGX) can be leveraged into an existing codebase to protect a security-sensitive application. Intel SGX provides user-level applications with hardware-enforced confidentiality and integrity protections and incurs manageable impact on performance. These protections apply to all three phases of the operational data lifecycle: at rest, in use, and in transit. SGX shrinks the trusted computing base (and therefore the attack surface) of the application to only the hardware on the CPU chip and the portion of the application's software that is executed within the protected enclave. The SDK enables SGX integration into existing C/C++ codebases while still ensuring program support for legacy and non-Intel platforms. This paper is the first published work to walk through the step-by-step process of Intel SGX integration with examples and performance results from an actual cryptographic application produced in a standard Linux development environment.