Michael Zhivich

Virtuoso: Narrowing the Semantic Gap in Virtual Machine Introspection

Introspection has featured prominently in many recent security solutions, such as virtual machine-based intrusion detection, forensic memory analysis, and low-artifact malware analysis. Widespread adoption of these approaches, however, has been hampered by the semantic gap: in order to extract meaningful information about the current state of a virtual machine, detailed knowledge of the guest operating systems inner workings is required. In this paper, we present a novel approach for automatically creating introspection tools for security applications with minimal human effort. By analyzing dynamic traces of small, in-guest programs that compute the desired introspection information, we can produce new programs that retrieve the same information from outside the guest virtual machine. We demonstrate the efficacy of our techniques by automatically generating 17 programs that retrieve security information across 3 different operating systems, and show that their functionality is unaffected by the compromise of the guest system. Our technique allows introspection tools to be effortlessly generated for multiple platforms, and enables the development of rich introspection-based security applications.

AWE: improving software analysis through modular integration of static and dynamic analyses

AWE is a prototype system for performing analysis of x86 executables in the absence of source code or debugging information. It provides a modular infrastructure for integrating static and dynamic analyses into a single workflow. One of the major challenges with performing analysis of modern software is the amount of data that must be analyzed by a human to determine software behavior. This challenge is further compounded by the number of different tools and extensive expertise required to perform such analyses. The AWE system addresses this challenge in two ways: first by focusing analysts attention on a prioritized subset of software features of importance, and second by simplifying analysis through an integrated static and dynamic analysis workflo.

You Sank My Battleship!: A Case Study in Secure Programming

We report on a case study in secure programming, focusing on the design, implementation and auditing of programs for playing the board game Battleship. We begin by precisely defining the security of Battleship programs, borrowing ideas from theoretical cryptography. We then consider three implementations of Battleship: one in Concurrent ML featuring a trusted referee; one in Haskell/LIO using information flow control to avoid needing a trusted referee; and one in Concurrent ML using access control to avoid needing such a referee. All three implementations employ data abstraction in key ways.

Securing Current and Future Process Control Systems

Process control systems (PCSs) are instrumental to the safe, reliable and efficient operation of many critical infrastructure components. However, PCSs increasingly employ commodity information technology (IT) elements and are being connected to the Internet. As a result, they have inherited IT cyber risks, threats and attacks that could affect the safe and reliable operation of infrastructure components, adversely affecting human safety and the economy. This paper focuses on the problem of securing current and future PCSs, and describes tools that automate the task. For current systems, we advocate specifying a policy that restricts control network access and verifying its implementation. We further advocate monitoring the control network to ensure policy implementation and verify that network use matches the design specifications. For future process control networks, we advocate hosting critical PCS software on platforms that tolerate malicious activity and protect PCS processes, and testing software with specialized tools to ensure that certain classes of vulnerabilities are absent prior to shipping.

Building low-power trustworthy systems: Cyber-security considerations for Real-Time Physiological Status Monitoring

Real-time monitoring of physiological data can reduce the likelihood of injury in noncombat military personnel and first-responders. MIT Lincoln Laboratory is developing a tactical Real-Time Physiological Status Monitoring (RT-PSM) system architecture and reference implementation named OBAN (Open Body Area Network), the purpose of which is to provide an open, government-owned framework for integrating multiple wearable sensors and applications. The OBAN implementation accepts data from various sensors enabling calculation of physiological strain information which may be used by squad leaders or medics to assess the teams health and enhance safety and effectiveness of mission execution. Security in terms of measurement integrity, confidentiality, and authenticity is an area of interest because OBAN system components exchange sensitive data in contested environments. In this paper, we analyze potential cyber-security threats and their associated risks to a generalized version of the OBAN architecture and identify directions for future research. The threat analysis is intended to inform the development of secure RT-PSM architectures and implementations.

Simulation Driven Policy Recommendations for Code Diversity

Periodic randomization of a computer program’s binary code is an attractive technique for defending against several classes of advanced threats. In this paper we describe a model of attacker-defender interaction in which the defender employs such a technique against an attacker who is actively constructing an exploit using Return Oriented Programming (ROP). In order to successfully build a working exploit, the attacker must guess the locations of several small chunks of program code, known as gadgets, in the defended program’s memory space. The defender thwarts the attacker’s efforts by periodically re-randomizing his code. Randomization incurs some performance cost, therefore an ideal strategy strikes an acceptable balance between utility degradation (cost) and security (benefit). We present risk aware and risk agnostic policy recommendations that were generated using simulation techniques. We found that policies that create low volatility environments are ideal for risk sensitive actors while policies that favor high system performance are more suitable for higher risk appetites.