Search for: All records

Although much of the work in behaviorally detecting malware lies in collecting the best explanatory data and using the most efficacious machine learning models, the processing of the data can sometimes prove to be the most important step in the data pipeline. In this work, we collect kernel-level system calls on a resource-constrained Internet of Things (IoT) device, apply lightweight Natural Language Processing (NLP) techniques to the data, and feed this processed data to two simple machine learning classification models: Logistic Regression (LR) and a Neural Network (NN). For the data processing, we group the system calls into n-grams that are sorted by the timestamp in which they are recorded. To demonstrate the effectiveness, or lack thereof, of using n-grams, we deploy two types of malware onto the IoT device: a Denial-of-Service (DoS) attack, and an Advanced Persistent Threat (APT) malware. We examine the effects of using lightweight NLP on malware like the DoS and the stealthy APT malware. For stealthier malware, such as the APT, using more advanced, but far more resource-intensive, NLP techniques will likely increase detection capability, which is saved for future work. 

We present research in the modeling of neurons within Drosophila (fruit fly) olfaction. We describe the process from data collection, to model creation, and spike generation. Our approach utilizes computational elements such as spiking neural networks that employ leaky integrate-and-fire neurons with adaptive firing behavior that more closely mimick biological neurons. We describe the methods of several learning implementations in both software and hardware. Finally, we present both quantitative and qualitative results on learning spiking neural network models. 

Abstract We study a second order ensemble method for fast computation of an ensemble of magnetohydrodynamics flows at small magnetic Reynolds number. Computing an ensemble of flow equations with different input parameters is a common procedure for uncertainty quantification in many engineering applications, for which the computational cost can be prohibitively expensive for nonlinear complex systems. We propose an ensemble algorithm that requires only solving one linear system with multiple right‐hands instead of solving multiple different linear systems, which significantly reduces the computational cost and simulation time. Comprehensive stability and error analyses are presented proving conditional stability and second order in time convergent. Numerical tests are provided to illustrate theoretical results and demonstrate the efficiency of the proposed algorithm. 

Abstract Weakly nonlinear, bi‐periodic patterns of waves that propagate in the‐direction with amplitude variation in the‐direction are generated in a laboratory. The amplitude variation in the‐direction is studied within the framework of the vector (vNLSE) and scalar (sNLSE) nonlinear Schrödinger equations using the uniform‐amplitude, Stokes‐like solution of the vNLSE and the Jacobi elliptic sine function solution of the sNLSE. The wavetrains are generated using the Stokes‐like solution of vNLSE; however, a comparison of both predictions shows that while they both do a reasonably good job of predicting the observed amplitude variation in, the comparison with the elliptic function solution of the sNLSE has significantly less error when the ratio of‐wavenumber to the two‐dimensional wavenumber is less than about 0.25. For ratios between about 0.25 and 0.30 (the limit of the experiments), the two models have comparable errors. When the ratio is less than about 0.17, agreement with the vNLSE solution requires a third‐harmonic term in the‐direction, obtained from a Stokes‐type expansion of interacting, symmetric wavetrains. There is no evidence of instability growth in the‐direction, consistent with the work of Segur and colleagues, who showed that dissipation stabilizes the modulational instability. Finally, there is some extra amplitude variation in, which is examined via a qualitative stability calculation that allows symmetry breaking in that direction. 

Recently, the Whitham and capillary Whitham equations were shown to accurately modelthe evolution of surface waves on shallow water. In order to gain a deeper understanding of theseequations, we compute periodic, traveling-wave solutions for both and study their stability. Wepresent plots of a representative sampling of solutions for a range of wavelengths, wave speeds, waveheights, and surface tension values. Finally, we discuss the role these parameters play in the stabilityof these solutions. 

Several problems challenge mesoscopic imaging in the brain: 1) Difficulty with positioning high-NA objectives near the brain; 2) Creating a flat imaging window against the surface of the brain; 3) Adjusting the imaging window in the face of changes in swelling and pressure in the brain; 4) Preventing growth of dura and biofilms that obscure the imaging window; 5) Follow-on MRI imaging of the animal post-implantation. We propose here an ultra-large window radiolucent implant to address these issues. Our approach provides a 2 cm diameter window for non-human primates (NHPs) that regulates pressure and employs a stable, strong, and thin design. The system is mechanically modeled and stress-tested to achieve access to the brain by large objectives, with design features that allow for manual repositioning of the imaging lens. To optimize the distance between the objective and the brain, we prioritize a thin implant design. A strong radiolucent implant is created using PEEK plastic, a strong, thermoresistant and biostable material. We heighten strength of the chamber’s attachment to the skull by using titanium screws that are normal to the surface of the bone at each point. The implant design has several parts and contemplates a potential method to maintain pressure on the brain. This method uses an engineered silicone mount to maintain even pressure of the imaging window on the brain’s surface, despite brain motion. The mechanical properties of the silicone are manipulated to closely resemble that of brain tissue to be more biomimetic and act as a cushion for motion. This method also allows for themanual repositioning of the cover slip to create a flat imaging window. Lastly, our approach prevents dural growth by blocking the migration of migratory biofilm-forming cells; we hypothesize that use of dynamic pressure maintenance on the brain is key to this method’s success. We are also investigating methods to elongate the longevity of the implant and imaging site, such as silver sputtering on implants and blue light therapy. These methods have produced an ultra-large field of view with 2P image results in <60,000 neurons. As such the chambers are expected to enhance recording window longevity and may prove to be a critical advance in NHP and human brain imaging.