No description available.

This dissertation introduces Power Fingerprinting (PFP), a novel technique for assessing the execution integrity of embedded devices. A PFP monitor is an external device that captures the dynamic power consumption of a processor using fine-grained measurements at the clock-cycle level and applies anomaly detection techniques to determine whether the integrity of the system has been compromised. PFP uses a set of trusted signatures from the target code that are extracted during a pre-characterization process. PFP provides significant visibility into the internal execution status, making it extremely robust against evasion. Because of its independence and physical separation, PFP prevents attacks on the monitor itself and introduces minimal overhead on platforms with resource constraints. Due to its anomaly detection operation, PFP is effective against unknown (zero-day) attacks. This dissertation demonstrates the feasibility of PFP on different platforms with different configurations and architectural complexities. Experimental results demonstrate the feasibility of PFP in a basic deterministic embedded platform for radio applications in two different areas: security and regulatory certification. For more complex, non-deterministic platforms, this works presents feasibility results for monitoring the execution integrity of complex software on a high-performance Android platform, including the ability to detect a real privilege escalation attack. In addition, the dissertation develops several general techniques to implement and integrate PFP into embedded platforms such as a general monitoring architecture, a methodology to characterize software modules and extract signatures, and an approach to perform board characterization and improve monitoring sensitivity. / Ph. D.

Mosquito pathogenic strains of Bacillus sphaericus are indistinguishable from nonpathogenic strains based on simple phenotypic tests. DNA-DNA hybridizations performed in 1980 placed the 7 pathogens included in that study in a distinct homology group separate from 5 groups of nonpathogens. The overall homology of the pathogenic strains to the species type strain was only 19% indicating that these pathogens should be a separate species. Since the DNA homology study was published in 1980, many more pathogenic strains have been isolated worldwide. Pathogenic strains have been differentiated from other strains of B. sphaericus by rRNA sequencing, fatty acid analysis, and isozyme analysis. The pathogens have been further classified by type of toxin produced, serotyping, and phage typing. I have used random amplified polymorphic DNA (RAPD) fingerprinting to determine the phenetic relationships among 31 pathogenic and 14 nonpathogenic strains of B. sphaericus. DNA Bands in agarose gel migrating the same distance were verified as being homologous using PCR-generated probes made from the RAPD bands. Band patterns resulting from 8 10-mer primers were examined by three coefficients, Jaccard, Dice, and simple matching. Each coefficient was able to distinguish DNA homology groups, although the relative similarity values differed. In agreement with DNA homology studies, pathogenic strains showed less than 10% similarity to nonpathogens using Jaccard and Dice coefficients. This value was 68% based on the simple matching coefficient. Individual serotypes were clearly indicated among the pathogenic strains by each coefficient. This suggests an overall genetic homogeneity among strains within serotypes. It also parallels the uniform toxicity pattern found within each serotype (unlike the toxin diversity found within B. thuringiensis serotypes). These results together with DNA homology data support the establishment of a new species for the pathogenic strains. / Master of Science

Over the last decade, there has been a tremendous growth in the number of electronic devices and applications. One of the very important aspects to deal with such proliferation of ICs is their security. Establishing the Identity (ID) of a device is the cornerstone of any secure application. Typically, the IDs of devices are stored in non-volatile memories (NVM) or through burning fuses on ICs. However, through such traditional techniques, IDs are vulnerable to attacks. Further, maintaining such secrets in NVMs is expensive. Physical Unclonable Functions (PUF) provide an alternative method for creating chip IDs. They exploit the uncontrollable variations that exist in IC manufacturing to generate identifiers. However, since PUFs exploit the small mismatch across identically designed circuits, the responses of PUFs are prone to error in the presence of unwanted variations in the operating temperature, supply voltage, and other noises. The overarching goal of this work is to develop silicon PUFs that are highly efficient and stable to such noises. In addition, to make PUFs more attractive for low cost and tiny embedded systems, our goal is to develop PUFs with minimal area and power consumption for a given ID length and security requirement. Techniques to develop such PUFs span different abstraction levels ranging from technology-independent application-level techniques to technology-dependent device-level ones. In this dissertation, we present different technology-independent and technology-dependent techniques and evaluate which techniques are good candidates for improving different qualities of PUFs. In technology-independent techniques, we propose two modifications to a conventional PUF architecture, which are detailed in this thesis. Both modifications result in a PUF that is more efficient in terms of area and power. Compared to the traditional architecture, for a given silicon real estate, the proposed architecture provides over two orders of magnitude larger $C/R$ space and it has higher resistance toward modeling attacks. Under technology-dependent methods, we investigate multiple techniques that improve stability and efficiency of PUF designs. In one approach, we propose a novel PUF design with a similar architecture to that of a traditional design, where we replace large and power hungry digital components with more efficient analog components. In another technique, we exploit the differences between pMOS and nMOS transistors in their variation of threshold voltage (Vth) and in the temperature coefficients of Vth to significantly improve the stability of bi-stable PUFs. We also use circuit-level simulations to evaluate the stability of silicon PUFs to aging degradation. We believe that our technology-independent techniques are good candidates for improving overall efficiency of PUFs in terms of both operation and implementation costs, suitable for PUFs with tight constraints on cost for design and test. However, with regards to improving the stability of PUFs, it is cost-effective to use our technology-dependent techniques as long as the extra effort for implementation and testing can be tolerated. / Ph. D.

Radio frequency fingerprinting, a type of physical layer identification, allows identifying wireless transmitters based on their unique hardware. Every wireless transmitter has slight manufacturing variations and differences due to the layout of components. These are manifested as differences in the signal emitted by the device. A variety of techniques have been proposed for identifying transmitters, at the physical layer, based on these differences. This has been successfully demonstrated on a large variety of transmitters and other devices. However, some situations still pose challenges: Some types of fingerprinting feature are very dependent on the modulated signal, especially features based on the frequency content of a signal. This means that changes in transmitter configuration such as bandwidth or modulation will prevent wireless fingerprinting. Such changes may occur frequently with cognitive radios, and in dynamic spectrum access networks. A method is proposed to transform features to be invariant with respect to changes in transmitter configuration. With the transformed features it is possible to re-identify devices with a high degree of certainty. Next, improving performance with limited data by identifying devices using observations crowdsourced from multiple receivers is examined. Combinations of three types of observations are defined. These are combinations of fingerprinter output, features extracted from multiple signals, and raw observations of multiple signals. Performance is demonstrated, although the best method is dependent on the feature set. Other considerations are considered, including processing power and the amount of data needed. Finally, drift in fingerprinting features caused by changes in temperature is examined. Drift results from gradual changes in the physical layer behavior of transmitters, and can have a substantial negative impact on fingerprinting. Even small changes in temperature are found to cause drift, with the oscillator as the primary source of this drift (and other variation) in the fingerprints used. Various methods are tested to compensate for these changes. It is shown that frequency based features not dependent on the carrier are unaffected by drift, but are not able to distinguish between devices. Several models are examined which can improve performance when drift is present. / Doctor of Philosophy / Radio frequency fingerprinting allows uniquely identifying a transmitter based on characteristics of the signal it emits. In this dissertation several extensions to current fingerprinting techniques are given. Together, these allow identification of transmitters which have changed the signal sent, identifying using different measurement types, and compensating for variation in a transmitter's behavior due to changes in temperature.

No

Yes / Pacific salmon (Oncorhynchus spp.) are ecological and cultural keystone species along the Northwest Coast of North America and are ubiquitous in archaeological sites of the region. The inability to morphologically identify salmonid post-cranial remains to species, however, can limit our understanding of the ecological and cultural role different taxa played in the seasonal subsistence practices of Indigenous groups in the past. Here, we present a rapid, cost-effective ZooMS method to distinguish salmonid species based on collagen peptide mass-fingerprinting. Using modern reference material and an assemblage of 28 DNA-identified salmonid bones from the pre-contact Yup'ik site of Nunalleq, Western Alaska, we apply high-resolution mass spectrometry (LC-MS/MS) to identify a series of potential collagen peptide markers to distinguish Pacific salmon. We then confirm these peptide markers with a blind ZooMS analysis (MALDI-TOF-MS) of the archaeological remains. We successfully distinguish five species of anadromous salmon with this ZooMS approach, including one specimen that could not be identified through ancient DNA analysis. Our biomolecular identification of chum (43%), sockeye (21%), chinook (18%), coho (11%) and pink (7%), confirm the exploitation of all five available species of salmonid at Nunalleq.

Radio Frequency Machine Learning (RFML) techniques for the classification tasks of Specific Emitter Identification (SEI) and Automatic Modulation Classification (AMC) have seen rapid improvements in recent years. The applications of SEI, a technique used to associate a received signal to an emitter, and AMC, a technique for determining the modulation scheme present within a transmission, are necessary for a variety of defense applications such as early warning systems and emitter tracking. Existing works studying SEI and AMC have sought to perform and improve classification through the use of various different machine learning (ML) model architectures. In ideal conditions, these efforts have shown strong classification results, however, when robust real-world data is applied to these models, performance notably decreases. Further efforts, therefore, are required to understand why each of these models fails in adverse conditions. With this understanding, robust architectures that are able to maintain performance in the presence of various data conditions can be created. The work presented in this thesis seeks to improve upon SEI and AMC models by furthering the understanding of how certain model architectures fail under varying data conditions, then applies Transfer Learning (TL) and Ensemble Learning techniques in an effort to mitigate discovered failures and improve the applicability of trained models to various types of data. Each of the approaches presented in this work utilize real-world datasets, collected in a way that emulate a variety of possible real life use conditions of RFML systems. Results show that existing AMC approaches are fairly robust to varying data conditions, while SEI approaches suffer a significant degradation in performance under conditions that differ than that used to train a given model. Further, TL and ensemble techniques can be utilized to improve the robustness of RFML models. This thesis helps isolate the rate and features of those SEI degradations, hopefully setting a foundation for future improvements. / Master of Science / Radio Frequency (RF) signals are produced by many different emitters encountered on a daily basis, including phones, networks, radar, and radios. These signals are used to transfer information from an emitter to a receiver, and contain a plethora of information that need be protected for defense practices in the RF domain. On the other hand, the information contained in these signals can be intercepted and utilized to discover information about potentially malicious transmissions. Two practices to determine information about received signals include Specific Emitter Identification (SEI), which relates an emitter to a received signal, and Automatic Modulation Classification (AMC), which determines the modulation scheme in which a signal is transmitted. A signal is made up of information, expressed in bits, and a modulation scheme is the method used to map those bits to express information. In recent years, Machine Learning (ML) techniques have been applied to SEI and AMC in an effort to improve the efficiency and accuracy results of classification. These ML approaches have shown high accuracy results when applied to data that is collected in the same environment as that used for training. When applied to data with different variables, however, model accuracy notably drops. This performance decrease motivates the need to discover more variables that negatively impact model performance, and further to create models that do not suffer from the same weaknesses. This work examines four different real-world variables that are common in deployed radio frequency machine learning (RFML) usage environments, and using the information learned about model failures, implements two approaches to create models that are more robust to variances in data. This work finds that model performance varies when exposed to variations in temperature, signal-to-noise ratio (SNR), training data quantity, and receiver hardware. Further, this work finds that Transfer Learning (TL) and Ensemble Learning can be used to create models that mitigate these discovered weaknesses.

Horses have been domesticated for thousands of years. Through selection practices horses have been separated into groups that pass on desired traits. We have viewed what the breeders have done at the genotypic level to accomplish their breeds. By using a relatively new technique, DNA fingerprinting, Thoroughbred inbreeding and eight other breeds (Standardbreds, Quarter horses, American Saddlebreds, National Show Horses, Arabians, Morgans, Mustangs and Belgians) have been studied. The probes CAC5 and YNZ 132 gave the best probability (ranging from 1.9x10⁻¹⁰ to 4.8x10⁻¹⁵) that two unrelated individuals would not have the same DNA fingerprint out of the probes screened. The level of inbreeding in Thoroughbreds has been estimated by comparing the number of bands shared among these animals to a quasi-natural population (Mustang) and a theoretically known genetic relationship (a sire and his offspring). Using the probes CAC5 and YNZ 132 Thoroughbreds share 20% more bands than the Mustang and 30-50% less than the sire and offspring. To compare the nine breeds, blood from ten horses from nine different breeds was mixed and DNA fingerprinted. Each lane on the autoradiograph therefore represents one breed. The two probes produced data with a rank correlation of .75 (Kendall's tau) (Ostle,B., 1963). Selection practices have been divided into, narrow selection regimes (where one or two traits have been selected for) and broad selection regimes (where numerous traits have been selected for). The amount of bands shared between the breeds was calculated and applied to a computer program named Gendiv (Gentzbittel and Nicolas, 1989,1991). Three consensus trees were derived showing that the narrow selection regime breeds, Thoroughbreds and Standardbreds, were the most genetically distanced from broad selection regime breeds, Mustangs and Morgans. / Master of Science

Positioning systems enable location-awareness for mobile devices, computers, and even tactical radios. From the collected location information, location-based services can be realized. One type of positioning system is based on location fingerprints. Unlike the conventional positioning techniques of time of or time delay of arrival (TOA/TDOA) or even angle of arrival (AOA), fingerprinting associates unique characteristics such as received signal strength (RSS) that differentiates a location from another location. The location-dependent characteristics then can be used to infer a user's location. Furthermore, fingerprinting requires no specialized hardware because of its reliance on an existing communications infrastructure. In estimating a user's position, fingerprint-based positioning systems are centrally calculated on a mobile computer using either a Euclidean distance algorithm, Bayesian statistics, or neural networks. With large service areas and, subsequently, large radio maps, one mobile computer may not have the adequate resources to locally compute a user's position. Wireless distributed computing provides a means for the mobile computer to meet the location-based service requirements and increase its network lifetime. This thesis develops distributed localization algorithms to be used in an indoor fingerprint-based positioning system. Fingerprint calculations are not computed on a single device, but rather on a wireless distributed computing network on Virginia Tech's Cognitive Radio Network Testbed (CORNET). / Master of Science