Search

In this dissertation, we present an information hiding approach incorporating anonymity that builds on existing classical steganographic models. Current security definitions are not sufficient to analyze the proposed information hiding approach as steganography offers data privacy by hiding the existence of data, a property that is distinct from confidentiality (data existence is known but access is restricted) and authenticity (data existence is known but manipulation is restricted). Combinations of the latter two properties are common in analyses, such as Authenticated Encryption with Associated Data (AEAD), yet there is a lack of research on combinations with steganography. This dissertation also introduces the security definition of Authenticated Stegotext with Associated Data (ASAD), which captures steganographic properties even when there is contextual information provided alongside the hidden data. We develop a hierarchical framework of ASAD variants, corresponding to different channel demands. We present a real-world steganographic embedding scheme, Authenticated SteGotex with Associated tRansaction Data (ASGARD), that leverages a blockchain-based application as a medium for sending hidden data. We analyze ASGARD in our framework and show that it meets Level-4 ASAD security. Finally, we implement ASGARD on the Ethereum platform as a proof-of-concept and analyze some of the ways an adversary might detect our embedding activity by analyzing historical Ethereum data., Lieutenant, United States Navy, Approved for public release. Distribution is unlimited.

How can we measure and classify neural network robustness across differently distributed data to avoid misuse of machine learning tools? This thesis adopts several metrics to measure the initial condition robustness of feedforward neural networks, allowing the creators of such networks to measure and refine their robustness and performance. This could allow highly robust neural networks to be used reliably on untrained data distributions and prevent the use of less robust networks as a black box in a poor environment. We test this measurement of robustness on a series of differently sized neural networks trained to detect and classify microgrid power system faults, giving examples of both robust and nonrobust networks, along with suggestions on how to maximize robustness. The analysis reveals that collecting data from segments along trajectories enhances the robustness of neural networks. In such data sets, the distribution of data points is dominated by the dynamics of the system, not the initial state distribution.

The applications of machine learning are broad and useful for a variety of industry and military objectives, but are the current methods robust? Robustness requires more than accuracy in ideal conditions; it means the system is resistant to perturbations in the data, both from natural and adversarial causes. This research aims to analyze the robustness of neural networks used for power-grid fault classifications. We focus on data generated from simulations of the classical 9-bus model; however, these methods and results can be extended to more complex microgrids, such as those found on naval ships, submarines, and bases. First, we measure the effects of random and adversarial noise on the testing data and compare three network types. Then we test different structures by varying the number of nodes and layers. Finally, we test whether adding noise to the training data can improve robustness. Before machine learning methods are adopted on submarines, we must first understand their weaknesses and potential for error. This research provides the foundation for how to test robustness, where neural networks are at risk from random or adversarial noise, and how to modify networks to improve their robustness.

This dissertation investigates a newly defined cryptographic differential, called a c-differential, and its relevance to the nonlinear substitution boxes of modern symmetric block ciphers. We generalize the notions of perfect nonlinearity, bentness, and avalanche characteristics of vectorial Boolean and p-ary functions using the c-derivative and a new autocorrelation function, while capturing the original definitions as special cases (i.e., when c=1). We investigate the c-differential uniformity property of the inverse function over finite fields under several extended affine transformations. We demonstrate that c-differential properties do not hold in general across equivalence classes typically used in Boolean function analysis, and in some cases change significantly under slight perturbations. Thus, choosing certain affine equivalent functions that are easy to implement in hardware or software without checking their c-differential properties could potentially expose an encryption scheme to risk if a c-differential attack method is ever realized. We also extend the c-derivative and c-differential uniformity into higher order, investigate some of their properties, and analyze the behavior of the inverse function's second order c-differential uniformity. Finally, we analyze the substitution boxes of some recognizable ciphers along with certain extended affine equivalent variations and document their performance under c-differential uniformity., Commander, United States Navy, Approved for public release. Distribution is unlimited.

A physical layer security framework is developed that may be tailored to a particular wireless communications link. Alternating space-time coding and Gray code hopping techniques are presented that may be implemented individually or concurrently. The first technique may be utilized in any multiple-input multiple-output system with choice of single- or multi-space-time code configurations. Orthogonal, non-orthogonal, and spatially-multiplexed space-time codes are included to achieve a desired diversity-multiplexing balance in addition to demonstrated security benefits. Gray code hopping may be used in any system with common digital modulation schemes to "hop" between alternative binary reflected Gray code mappings of bits to complex-valued modulation symbols. Unlike similar constellation remapping techniques, Gray code hopping provides improved physical layer security against plaintext attacks without sacrificing bit error rate performance. To facilitate these schemes, we present two series of algorithms that systematically build large sets of unique alternative space-time codes and modulation mappings. Decoding methods are discussed to allow for channel equalization and symbol co-phasing when employing an alternating space-time code. Equivalent modulation/demodulation approaches are provided for Gray code hopping. Information-theoretic and computational security analysis, Monte Carlo simulations, and over-the-air testing are completed as confirmation of proposed methods., Lieutenant Commander, United States Navy, Approved for public release. Distribution is unlimited.

Encryption is ubiquitous in the modern environment. While public/private key architecture has provided an amazing and powerful way to encrypt information so that only one intended recipient can decrypt, the computation required for this approach means that this encryption method can quickly grow extremely expensive. With that in mind, there are a variety of open-source stream ciphers that seek to provide relatively inexpensive stream ciphers to securely encrypt information. But these stream ciphers all operate using very different techniques to generate their keystream, as seen in the stark differences in paradigms between ciphers. As such, it is not immediately clear what operations are required to achieve the desired level of encryption. What cryptographic primitives are most common or efficacious in achieving security? Examining the Data Encryption Standard, Advanced Encryption Standard, and the stream cipher winners of the eStream II competition, an underlying trend composed of two operations emerges. Despite observing no clear n-grams defining precise cryptographic primitives, we identify a general structure common to all stream ciphers. Additionally, we identify that substitution boxes or multiplication operations are not necessary for stream ciphers, whereas addition and rotation operations seem to be essential., Major, United States Army, Approved for public release. Distribution is unlimited.

We discuss in this paper the independence numbers and algebraic properties of I-graphs. The I-graphs are a further generalization of the Generalized Petersen graphs whose independence numbers have been previously researched. Specifically, we give bounds for the independence number of different I-graphs and sub-classes of I-graphs, and exactly determine the independence number for other I-graphs and sub-classes of I-graphs. We also analyze the automorphism groups of the I-graphs. These groups have been characterized in previous papers; in this paper, we examine them via their Cayley graphs. These Cayley graphs are characterized completely and examined according to their graph theoretical and algebraic properties., Ensign, United States Navy, Approved for public release. distribution is unlimited

Cryptography is widely used by everybody in day-to-day activities. Many cryptographic algorithms rely on pseudorandom number sequences. One of the quickest methods of pseudorandom number generation is using linear feedback shift registers (LFSR) to generate sequences. LFSR sequences exhibit good statistical properties, but alone are not adequately secure due to their low linear complexity. To enhance the security, separate LFSR sequences can be combined into a single pseudorandom string by using a combiner. In this thesis, creating a combiner function using another pseudorandom sequence derived from the chaotic motion of a double pendulum is investigated. Using the information from this driving function, an iterative process occurs whereby certain LFSR sequence blocks are selected and combined. The resultant sequences are sufficiently random as proven by the 15 tests adopted by the National Institute of Standards and Technology to evaluate the randomness of binary strings.

Moving target defense (MTD) is a promising strategy for gaining advantage over cyber attackers, but these dynamic reconfigurations can impose significant overhead. We propose implementing MTD within an optimization framework so that we seize defensive advantage while minimizing overhead. This dissertation presents an MTD scheme that leverages partially observable Markov decision processes (POMDP) with absorbing states to select the optimal defense based on partial observations of the cyber attack phase. In this way, overhead is minimized as reconfigurations are triggered only when the potential benefit outweighs the cost. We formulate and implement a POMDP within a system with Monte-Carlo planning-based decision making configured to reflect defender-defined priorities for the cost-benefit tradeoff. The proposed system also includes a performance -monitoring scheme for continuous validation of the model, critical given attackers' ever-changing techniques. We present simulation results that confirm the system fulfills the design goals, thwarting 99% of inbound attacks while sustaining system availability at greater than 94% even as probability of attack phase detection dropped to 0.74. A comparable system that triggered MTD techniques pseudorandomly maintained just 43% availability when providing equivalent attack suppression, which illustrates the utility of our proposed scheme.

The double pendulum is a system of two connected masses, one tethered to a point in space and the other tethered to the first mass. The double pendulum exhibits chaotic motion under the influence of an external force such as the gravitational force. The chaotic motion is sensitive to the initial conditions or positions of the masses, resulting in an infinite number of possible motion paths. The chaotic motion paths produced by the double pendulum can be exploited to produce binary sequences. This thesis focuses on determining if the chaotic motion of the double pendulum can be used as a pseudorandom number generator (PRNG) by modeling the motion and using different methods of extracting bits from the motion paths to produce pseudorandom binary sequences. The pseudorandom binary sequences are then evaluated using tests for randomness as defined by the National Institute of Standards and Technology (NIST). The methods of bit extraction can then be compared based on their NIST test results to ultimately determine the practicality of the double pendulum as a PRNG. Considerations can be made for different methods of bit extraction or the use of different chaotic motions.

In this paper, we study the independence number of various classes of I-graphs. I-graphs are generalizations of the Generalized Petersen Graphs. We provide constructions of independent sets given any parameters, as well as bounds for some subclasses of I-graphs. We also prove exact results for some specific I-graphs., Ensign, United States Navy, Approved for public release. distribution is unlimited

The purpose of this thesis is to analyze the cryptographic properties of a pseudorandom bit generator that combines Blum Blum Shub and linear feedback shift register sequences using a shrinking generator configuration. We sought to answer the questions: (1) What are the strengths and weaknesses of this type of combiner? (2) What constraints must be placed on the input parameters to ensure good cryptographic properties of the output sequence? We generated sequences using variations of this combiner. We then evaluated their cryptographic suitability with the National Institute of Standards and Technology (NIST) statistical test suite. We identified lower bounds on the input parameters to increase the probability that the combiner would perform well under the NIST test suite. Our scheme produced consistently excellent results under NIST testing but is computationally too slow for many practical uses as a stream cipher. Future work could focus on methods to increase the speed of the generator without a loss of excellent cryptographic properties., Major, United States Army, Approved for public release. distribution is unlimited

Moving target defense (MTD) is a promising strategy for gaining advantage over cyber attackers, but these dynamic reconfigurations can impose significant overhead. We propose implementing MTD within an optimization framework so that we seize defensive advantage while minimizing overhead. This dissertation presents an MTD scheme that leverages partially observable Markov decision processes (POMDP) with absorbing states to select the optimal defense based on partial observations of the cyber attack phase. In this way, overhead is minimized as reconfigurations are triggered only when the potential benefit outweighs the cost. We formulate and implement a POMDP within a system with Monte-Carlo planning-based decision making configured to reflect defender-defined priorities for the cost-benefit tradeoff. The proposed system also includes a performance -monitoring scheme for continuous validation of the model, critical given attackers' ever-changing techniques. We present simulation results that confirm the system fulfills the design goals, thwarting 99% of inbound attacks while sustaining system availability at greater than 94% even as probability of attack phase detection dropped to 0.74. A comparable system that triggered MTD techniques pseudorandomly maintained just 43% availability when providing equivalent attack suppression, which illustrates the utility of our proposed scheme., Lieutenant Commander, United States Navy, Approved for public release. distribution is unlimited