USMA Digital Commons (United States Military Academy, West Point)
Not a member yet
1355 research outputs found
Sort by
Principles of Robust Learning and Inference for Internet of Battlefield Things (In Internet of Things for Defense and National Security)
No full textThe Internet of Battlefield Things (IoBTs) aims at providing a pervasive, heterogeneous sensing and actuation capability to enhance command and control system autonomy and agility, information analytic capabilities against adversarial influence and control of the information battle-space; delivering intelligent, agile, and resilient decisional overmatch at significant standoff and optempo. While the traditional approaches have focused on either centralized or decentralized decision-making, with the decision structure either fixed vertical stovepipes or dynamic task organized, and the information dissemination either limited (need to know) or broadcasted (need to share), IoBTs aim at providing options across these extremes of the spectrum to provide an adaptive mission-oriented network of sensors and actuators. Thus, the discovery, composition and adaptation of available network nodes for sensing, secure information sharing, and actuation is a critical capability for IoBTs. This has motivated enabling intelligent services as core components of IoBTs to make them autonomous and to enable services necessary for effective command and control. The examples of such artificial intelligence (AI) services that need to be supported by the complex autonomic IoBTs include intelligent analytics, anomaly detection in broadly heterogeneous and varied data that may be unknown combinations of sparse and voluminous, and centralized and distributed decision-making on whether received data is trustworthy or suspect. Further, the adversarial nature of the contested environment in which IoBTs operate requires enriching the resiliency of the IoBT, such that it can be hardened against tampering and adversarial compromise, continue operating under attacks, and provide bounded guarantees of performance. The tremendous success of machine learning, in particular deep learning methods, make them a promising paradigm to develop and deploy the intelligent services in an IoBT. But these machine learning models are known to be brittle, untrustworthy, and vulnerable to adversarial attacks. These limitations have fueled our research into principles and methodologies to make machine learning models robust, resilient to adversarial attacks, uncertainty-aware, and more interpretable for human-on-the-loop decision-making.https://digitalcommons.usmalibrary.org/aci_books/1021/thumbnail.jp
Constrained Optimization Based Adversarial Example Generation for Transfer Attacks in Network Intrusion Detection Systems
Full text linkDeep learning has enabled network intrusion detection rates as high as 99.9% for malicious network packets without requiring feature engineering. Adversarial machine learning methods have been used to evade classifiers in the computer vision domain; however, existing methods do not translate well into the constrained cyber domain as they tend to produce non-functional network packets. This research views the payload of network packets as code with many functional units. A meta-heuristic based generative model is developed to maximize classification loss of packet payloads with respect to a surrogate model by repeatedly substituting units of code with functionally equivalent counterparts. The perturbed packets are then transferred and tested against three test network intrusion detection system classifiers with various evasion rates that depend on the classifier and malicious packet type. If the test classifier is of the same architecture as the surrogate model, near-optimal adversarial examples penetrate the test model for 69% of packets whereas the raw examples succeeds for only 5% of packets. This confirms hypotheses that NIDS classifiers are vulnerable to adversarial attacks, motivating research in robust learning for cyber
ELVIS AND EMINEM: THE CULTURAL IMPLICATIONS OF AMERICAN MIDDLE-CLASS ASSIMILATION OF AFRICAN-AMERICAN MUSIC
Full text link
Prioritizing Ports and Waterways Safety Assessments (PAWSAs)
Full text linkA Ports and Waterways Safety Assessment (PAWSA) is a discussion forum facilitated by the United States Coast Guard Navigation Center (NAVCEN) to analyze the state of a port or waterway as it relates to navigation, vessel traffic, and physical attributes of the waterway. A successful workshop requires the collaboration between various stakeholders such as waterway users, environmental interest groups and local law enforcement. Without involving the relevant parties and encouraging their insight, the USCG risks creating an incomplete understanding of the port’s dynamic.
This project examines the characteristics of eight ports across the U.S. to determine which is in most need of a PAWSA. NAVCEN is limited to conducting three to five PAWSAs per year and seeks to maximize the efficacy of the workshops by traveling to the ports that need them the most. A customizable decision support tool based on quantitative factors was created to modernize NAVCEN’s current decision-making process for assigning ports PAWSAs. Tools such as ArcGIS Pro and Python were used