Publications
Topological effects on attacks against vertex classification
March 12, 2020
Journal Article
Published in:
https://arxiv.org/abs/2003.05822
Topic:
R&D area:
Summary
Vertex classification is vulnerable to perturbations of both graph topology and vertex attributes, as shown in recent research. As in other machine learning domains, concerns about robustness to adversarial manipulation can prevent potential users from adopting proposed methods when the consequence of action is very high. This paper considers two topological characteristics of graphs and explores the way these features affect the amount the adversary must perturb the graph in order to be successful. We show that, if certain vertices are included in the training set, it is possible to substantially an adversary's required perturbation budget. On four citation datasets, we demonstrate that if the training set includes high degree vertices or vertices that ensure all unlabeled nodes have neighbors in the training set, we show that the adversary's budget often increases by a substantial factor---often a factor of 2 or more---over random training for the Nettack poisoning attack. Even for especially easy targets (those that are misclassified after just one or two perturbations), the degradation of performance is much slower, assigning much lower probabilities to the incorrect classes. In addition, we demonstrate that this robustness either persists when recently proposed defenses are applied, or is competitive with the resulting performance improvement for the defender.
Summary
Vertex classification is vulnerable to perturbations of both graph topology and vertex attributes, as shown in recent research. As in other machine learning domains, concerns about robustness to adversarial manipulation can prevent potential users from adopting proposed methods when the consequence of action is very high. This paper considers two...
READ MORE
Augmented Annotation Phase 3
March 9, 2020
Technical Report
Published in:
MIT Lincoln Laboratory Report TR-1248
Summary
Automated visual object detection is an important capability in reducing the burden on human operators in many DoD applications. To train modern deep learning algorithms to recognize desired objects, the algorithms must be "fed" more than 1000 labeled images (for 55%–85% accuracy according to project Maven - Oct 2017 O6, Working Group slide 27) of each particular object. The task of labeling training data for use in machine learning algorithms is human intensive, requires special software, and takes a great deal of time. Estimates from ImageNet, a widely used and publicly available visual object detection dataset, indicate that humans generated four annotations per minute in the overall production of ImageNet annotations. DoD's need is to reduce direct object-by-object human labeling particularly in the video domain where data quantity can be significant. The Augmented Annotations System addresses this need by leveraging a small amount of human annotation effort to propagate human initiated annotations through video to build an initial labeled dataset for training an object detector, and utilizing an automated object detector in an iterative loop to assist humans in pre-annotating new datasets.
Summary
Automated visual object detection is an important capability in reducing the burden on human operators in many DoD applications. To train modern deep learning algorithms to recognize desired objects, the algorithms must be "fed" more than 1000 labeled images (for 55%–85% accuracy according to project Maven - Oct 2017 O6...
READ MORE
Medical countermeasures analysis of 2019-nCoV and vaccine risks for antibody-dependent enhancement (ADE)
March 8, 2020
Journal Article
Published in:
https://www.preprints.org/manuscript/202003.0138/v1
Topic:
R&D area:
R&D group:
Summary
Background: In 80% of patients, COVID-19 presents as mild disease. 20% of cases develop severe (13%) or critical (6%) illness. More severe forms of COVID-19 present as clinical severe acute respiratory syndrome, but include a T-predominant lymphopenia, high circulating levels of proinflammatory cytokines and chemokines, accumulation of neutrophils and macrophages in lungs, and immune dysregulation including immunosuppression. Methods: All major SARS-CoV-2 proteins were characterized using an amino acid residue variation analysis method. Results predict that most SARS-CoV-2 proteins are evolutionary constrained, with the exception of the spike (S) protein extended outer surface. Results were interpreted based on known SARS-like coronavirus virology and pathophysiology, with a focus on medical countermeasure development implications. Findings: Non-neutralizing antibodies to variable S domains may enable an alternative infection pathway via Fc receptor-mediated uptake. This may be a gating event for the immune response dysregulation observed in more severe COVID-19 disease. Prior studies involving vaccine candidates for FCoV SARS-CoV-1 and Middle East Respiratory Syndrome coronavirus (MERS-CoV) demonstrate vaccination-induced antibody-dependent enhancement of disease (ADE), including infection of phagocytic antigen presenting cells (APC). T effector cells are believed to play an important role in controlling coronavirus infection; pan-T depletion is present in severe COVID-19 disease and may be accelerated by APC infection. Sequence and structural conservation of S motifs suggests that SARS and MERS vaccine ADE risks may foreshadow SARS-CoV-2 S-based vaccine risks. Autophagy inhibitors may reduce APC infection and T-cell depletion. Amino acid residue variation analysis identifies multiple constrained domains suitable as T cell vaccine targets. Evolutionary constraints on proven antiviral drug targets present in SARS-CoV-1 and SARS-CoV-2 may reduce risk of developing antiviral drug escape mutants. Interpretation: Safety testing of COVID-19 S protein-based B cell vaccines in animal models is strongly encouraged prior to clinical trials to reduce risk of ADE upon virus exposure.
Summary
Background: In 80% of patients, COVID-19 presents as mild disease. 20% of cases develop severe (13%) or critical (6%) illness. More severe forms of COVID-19 present as clinical severe acute respiratory syndrome, but include a T-predominant lymphopenia, high circulating levels of proinflammatory cytokines and chemokines, accumulation of neutrophils and macrophages...
READ MORE
Toward an autonomous aerial survey and planning system for humanitarian aid and disaster response
March 7, 2020
Conference Paper
Published in:
IEEE Aerospace Conf., 7-14 March 2020.
Summary
In this paper we propose an integrated system concept for autonomously surveying and planning emergency response for areas impacted by natural disasters. Referred to as AASAPS-HADR, this system is composed of a network of ground stations and autonomous aerial vehicles interconnected by an ad hoc emergency communication network. The system objectives are three-fold: to provide situational awareness of the evolving disaster event, to generate dispatch and routing plans for emergency vehicles, and to provide continuous communication networks which augment pre-existing communication infrastructure that may have been damaged or destroyed. Lacking development in previous literature, we give particular emphasis to the situational awareness objective of disaster response by proposing an autonomous aerial survey that is tasked with assessing damage to existing road networks, detecting and locating human victims, and providing a cursory assessment of casualty types that can be used to inform medical response priorities. In this paper we provide a high-level system design concept, identify existing AI perception and planning algorithms that most closely suit our purposes as well as technology gaps within those algorithms, and provide initial experimental results for non-contact health monitoring using real-time pose recognition algorithms running on a Nvidia Jetson TX2 mounted on board a quadrotor UAV. Finally we provide technology development recommendations for future phases of the AASAPS-HADR system.
Summary
In this paper we propose an integrated system concept for autonomously surveying and planning emergency response for areas impacted by natural disasters. Referred to as AASAPS-HADR, this system is composed of a network of ground stations and autonomous aerial vehicles interconnected by an ad hoc emergency communication network. The system...
READ MORE
Automated discovery of cross-plane event-based vulnerabilities in software-defined networking
February 23, 2020
Conference Paper
Published in:
Network and Distributed Systems Security Symp., NDSS, 23-26 February 2020.
Summary
Software-defined networking (SDN) achieves a programmable control plane through the use of logically centralized, event-driven controllers and through network applications (apps) that extend the controllers' functionality. As control plane decisions are often based on the data plane, it is possible for carefully crafted malicious data plane inputs to direct the control plane towards unwanted states that bypass network security restrictions (i.e., cross-plane attacks). Unfortunately, because of the complex interplay among controllers, apps, and data plane inputs, at present it is difficult to systematically identify and analyze these cross-plane vulnerabilities. We present EVENTSCOPE, a vulnerability detection tool that automatically analyzes SDN control plane event usage, discovers candidate vulnerabilities based on missing event-handling routines, and validates vulnerabilities based on data plane effects. To accurately detect missing event handlers without ground truth or developer aid, we cluster apps according to similar event usage and mark inconsistencies as candidates. We create an event flow graph to observe a global view of events and control flows within the control plane and use it to validate vulnerabilities that affect the data plane. We applied EVENTSCOPE to the ONOS SDN controller and uncovered 14 new vulnerabilities.
Summary
Software-defined networking (SDN) achieves a programmable control plane through the use of logically centralized, event-driven controllers and through network applications (apps) that extend the controllers' functionality. As control plane decisions are often based on the data plane, it is possible for carefully crafted malicious data plane inputs to direct the...
READ MORE
Unified value-based feedback, optimization and risk management in complex electric energy systems
February 22, 2020
Journal Article
Published in:
Optim Eng 21, 427–483 (2020)
Summary
The ideas in this paper are motivated by an increased need for systematic data-enabled resource management of large-scale electric energy systems. The basic control objective is to manage uncertain disturbances, power imbalances in particular, by optimizing available power resources. To that end, we start with a centralized optimal control problem formulation of system-level performance objective subject to complex interconnection constraints and constraints representing highly heterogeneous internal dynamics of system components. To manage spatial complexity, an inherent multi-layered structure is utilized by modeling interconnection constraints in terms of unifed power variables and their dynamics. Similarly, the internal dynamics of components and sub-systems (modules), including their primary automated feedback control, is modeled so that their input–output characterization is also expressed in terms of power variables. This representation is shown to be key to managing the multi-spatial complexity of the problem. In this unifying energy/ power state space, the system constraints are all fundamentally convex, resulting in the convex dynamic optimization problem, for typically utilized quadratic cost functions. Based on this, an interactive multi-layered modeling and control method is introduced. While the approach is fundamentally based on the primal–dual decomposition of the centralized problem, this is formulated for the frst time for the couple real-reactive power problem. It is also is proposed for the frst time to utilize sensitivity functions of distributed agents for solving the primal distributed problem. Iterative communication complexity typically required for convergence of pointwise information exchange is replaced by the embedded distributed optimization by the modules when creating these functions. A theoretical proof of the convergence claim is given. Notably, the inherent multi-temporal complexity is managed by performing model predictive control (MPC)-based decision making when solving distributed primal problems. The formulation enables distributed decision-makers to value uncertainties and related risks according to their preferences. Ultimately, the distributed decision making results in creating a bid function to be used at the coordinating market-clearing level. The optimization approach in this paper provides a theoretical foundation for next-generation Supervisory Control and Data Acquisition (SCADA) in support of a Dynamic Monitoring and Decision Systems (DyMonDS) for a multi-layered interactive market implementation in which the grid users follow their sub-objectives and the higher layers coordinate interconnected sub-systems and the high-level system objectives. This forms a theoretically sound basis for designing IT-enabled protocols for secure operations, planning, and markets.
Summary
The ideas in this paper are motivated by an increased need for systematic data-enabled resource management of large-scale electric energy systems. The basic control objective is to manage uncertain disturbances, power imbalances in particular, by optimizing available power resources. To that end, we start with a centralized optimal control problem...
READ MORE
Wind information requirements for NextGen applications phase 7 report
February 12, 2020
Project Report
Published in:
MIT Lincoln Laboratory Report ATC-444
Topic:
R&D area:
Summary
This report details the Required Time of Arrival (RTA) performance of B757 aircraft arriving at various meter fixes across a range of altitudes from 33,000' down to 3,000' above ground level (AGL). The system tested demonstrated less than ±10 second arrival error in at least 95% of flights at meter fixes down to 7,000' AGL regardless of the forecast quality provided. Below 7,000' AGL, RTA performance significantly degraded demonstrating around 80% compliance under the best forecast and operating conditions. This report also provides a comprehensive lexicon of aviation and air traffic control related "wind" terms.
Summary
This report details the Required Time of Arrival (RTA) performance of B757 aircraft arriving at various meter fixes across a range of altitudes from 33,000' down to 3,000' above ground level (AGL). The system tested demonstrated less than ±10 second arrival error in at least 95% of flights at meter...
READ MORE
Safe predictors for enforcing input-output specifications [e-print]
January 29, 2020
Journal Article
Published in:
https://arxiv.org/abs/2001.11062
Summary
We present an approach for designing correct-by-construction neural networks (and other machine learning models) that are guaranteed to be consistent with a collection of input-output specifications before, during, and after algorithm training. Our method involves designing a constrained predictor for each set of compatible constraints, and combining them safely via a convex combination of their predictions. We demonstrate our approach on synthetic datasets and an aircraft collision avoidance problem.
Summary
We present an approach for designing correct-by-construction neural networks (and other machine learning models) that are guaranteed to be consistent with a collection of input-output specifications before, during, and after algorithm training. Our method involves designing a constrained predictor for each set of compatible constraints, and combining them safely via...
READ MORE
AI data wrangling with associative arrays [e-print]
January 18, 2020
Abstract
Published in:
Submitted to Northeast Database Day, NEDB 2020, https://arxiv.org/abs/2001.06731
Topic:
R&D area:
R&D group:
Summary
The AI revolution is data driven. AI "data wrangling" is the process by which unusable data is transformed to support AI algorithm development (training) and deployment (inference). Significant time is devoted to translating diverse data representations supporting the many query and analysis steps found in an AI pipeline. Rigorous mathematical representations of these data enables data translation and analysis optimization within and across steps. Associative array algebra provides a mathematical foundation that naturally describes the tabular structures and set mathematics that are the basis of databases. Likewise, the matrix operations and corresponding inference/training calculations used by neural networks are also well described by associative arrays. More surprisingly, a general denormalized form of hierarchical formats, such as XML and JSON, can be readily constructed. Finally, pivot tables, which are among the most widely used data analysis tools, naturally emerge from associative array constructors. A common foundation in associative arrays provides interoperability guarantees, proving that their operations are linear systems with rigorous mathematical properties, such as, associativity, commutativity, and distributivity that are critical to reordering optimizations.
Summary
The AI revolution is data driven. AI "data wrangling" is the process by which unusable data is transformed to support AI algorithm development (training) and deployment (inference). Significant time is devoted to translating diverse data representations supporting the many query and analysis steps found in an AI pipeline. Rigorous mathematical...
READ MORE
Beamforming with distributed arrays: FY19 RF Systems Line-Supported Program
January 7, 2020
Project Report
Published in:
MIT Lincoln Laboratory Report LSP-270
Topic:
R&D area:
R&D group:
Summary
Spatial beamforming using distributed arrays of RF sensors is treated. Unlike the observations from traditional RF antenna arrays, the distributed array's data can be subjected to widely varying time and frequency shifts among sensors and signals. These shifts require compensation upon reception in order to perform spatial filtering. To perform beamforming with a distributed array, the complex-valued observations from the sensors are shifted in time and frequency, weighted, and summed to form a beamformer output that is designed to mitigate interference and enhance signal energy. The appropriate time-frequency shifts required for good beamforming are studied here using several different methodologies.
Summary
Spatial beamforming using distributed arrays of RF sensors is treated. Unlike the observations from traditional RF antenna arrays, the distributed array's data can be subjected to widely varying time and frequency shifts among sensors and signals. These shifts require compensation upon reception in order to perform spatial filtering. To perform...
READ MORE