United States

In our research, we successfully created a custom neural trojan by modifying the weights of the model itself. Additionally, we successfully defended against this attack by adding a TensorFlow Extension to encrypt and hash custom checkpoints used in the training and testing of the model. 


NAML 2021

Designing Secure Machine Learning Systems

neural trojan

supply chain

machine learning

The purpose of the AAAI conference is to promote research in artificial intelligence (AI) and scientific exchange among AI researchers, practitioners, scientists, and engineers in affiliated disciplines. 

AAAI 2021

poster

Naval Information Warfare Center Pacific is partnering with AFCEA San Diego chapter to host the workshop.

The NAML workshop is an opportunity for researchers from Navy research and development centers and our partner organizations to exchange technical ideas and to come together as a team.

The goals of NAML are:

* To spread awareness of current machine learning research relevant to Naval applications

* To connect machine learning researchers with experts in military requirements

* To build and strengthen collaborations within the defense research community

**Watch this video to hear about the AFCEA purpose.**

<iframe width=580 height=700 frameborder="0" scrolling="no" src="https://screencast-o-matic.com/player/creleVVcvRh?v=6&ff=1&title=0&controls=1" allowfullscreen="true"></iframe>

Please log in to explore this event. You need to log in with the email address you registered with. Access credentials have been sent to your email. 

Please be sure to check your spam and other email folders if you do not see an email confirmation right away.

Please log in to explore this event.

Registration for this workshop has now closed. We appreciate your interest and hope to welcome you at a future event. If you’d like to hear about similar upcoming events, please check our website or join our mailing list.

Registration for this workshop has now closed.

The annual workshop on Naval Applications of Machine Learning is an opportunity for researchers from Navy research and development centers and our partner organizations to exchange technical ideas and to come together as a team. The goals of NAML are to spread awareness of current machine learning research relevant to Naval applications, to connect machine learning researchers with experts in military requirements and to build and strengthen collaborations within the defense research community.


Predictive analytics offers a potential game changing capability for Naval tactical decision superiority. Tactical operations could take a significant leap in progress with the aid of a real-time automated predictive analytics capability that provides predictions of second and third order effects of possible courses of action. This future capability would accompany current developments in the use of artificial intelligence and data analytics to improve battlespace knowledge and offer automated battle management aids to the tactical warfighter. As the automated battle management aids develop tactical course of action options the predictive analytics capability could predict how the adversary might respond to each course of action option. The predictive analytics capability could continue to “wargame” possible blue force/red force actions and responses—generating predictions of second and third order effects. These predictions offer the tactical warfighter a more strategic perspective in making tactical course of action decisions. By performing this analysis using an automated aid with artificial intelligence, it allows the capability to sup-port real-time decisions and to analyze great amounts of data (both sensor data and historical data) and handle highly complex tactical environments This real-time wargaming translates into high order computations that would be impossible to be performed manually in the short reaction times given. This paper discusses the results of a study of predictive analytic capabilities in the naval maritime do-main.

Predictive Analytics in the Maritime Domain

In recent years, there have been significant parallel advances in two areas: laser systems and artificial intelligence (AI). The Navy is rapidly moving forward with the development and deployment of laser systems as weapons and sensors. In order to realize their full potential, laser systems must adapt to their environment and be utilized in concert with existing kinetic weapons and tactical sensors for effective engagements and military operations. Decision-making regarding the use of laser systems will be challenging in complex environments. AI methods, and automation in general, including machine learning, data analytics, and predictive analytics have the potential to create a “cognitive” laser that can assess a threat situation in real-time and adapt itself to overcome environmental challenges, select a specific target aimpoint, and tailor its beam to specific target characteristics.

The Naval Postgraduate School (NPS) is developing concepts for a cognitive laser to support the operation of shipboard laser weapon systems in the maritime domain. NPS is conducting a model-based systems engineering (MBSE) approach to understanding the laser system’s complex decision space, studying human-machine teaming methods, developing and analyzing models of decision aid concepts, and performing proof-of-concept demonstrations of machine learning methods. This presentation will cover the NPS research team approach as well as discussing concepts, progress, and preliminary analysis results.

Cognitive Laser – Automated Decision Aid for Laser Weapon Systems


You click the screen on your phone to enter your banking app, confirm your checking balance, and then you put the phone away. This simple interaction has been captured by not only your banking app, but other apps that are currently active on your phone. Data is collected on how long you interact with different screens, how quickly you click on features, and even covertly captured photos of you to classify whether you appear happy or unemotional while looking at data. This information is used to improve system functionality, allowing apps to become easier to use. This Big Data collection happens on nearly every phone, and has become so common that we confirm sensitive data collection every time we download a new app. As we have learned over time, there is little reason to worry about our individual data. Companies use the information to confirm changes amongst large amounts of users, further improving their software and user interfaces. 

Why doesn’t the Navy collect large swaths of data to analyze not only user inputs, but other features that are arguably more important to our warfighters’ wellbeing? The Navy has hundreds of hardware and software used by our warfighters on a daily basis, but information is either not collected or goes into a black box that only the developer can access. Our warfighters frequently interact with interfaces that are extremely different from one another on a daily basis, with some interfaces taking 4 hours or longer to interact with and carry out functions, in addition to standard work duties. These interactions could have been developed to be streamlined and similar if only the data was easily accessible.

In order to bring this capability to the forefront in the Navy, we are collecting information and user interaction data from robotic craft and control systems in an attempt to better understand systems and user dynamics. This data collection will allow for improvements to system resilience, automation, and user experience interactions on the user interface. By taking this information and pairing it with big data techniques, artificial intelligence (AI), and machine learning (ML), we can enable quicker information dissemination to our warfighters, allowing them to carry on with the task at hand. With this data, we can identify common features and streamline uses amongst hundreds of different project types. User interactions can be recorded on a larger scale and used across dissimilar systems to ensure ease of use, much like what can be seen in phone apps. As Navy increases data sharing amongst efforts, systems will become more efficient, increasing the speed with which we send vital technologies to the Fleet, and ensuring our systems are streamlined for our warfighters.

Navy Big Data, Bringing Software Consistency to the Fleet

Historically, recorded Naval F/A-18 aircraft status data has been sparse. With more populated recent records, aircraft status and state transitions have become an opportunity for data mining and cluster analysis. Aircraft state transition segmentation analysis provides high level insights into readiness, mission capability (MC), and return-on-investment. Event-based temporal analysis provides insights into how events (e.g. maintenance) affect aircraft and their movement between clusters. In this effort, aircraft-level state transitions are segmented by k-means clustering. Furthermore, event-based temporal analysis of aircraft clusters was performed through statistical methods. Results of this model assist decision makers and support the Digital Aviation Readiness Technology Engine (DARTE) family of ML/AI models.

State Transition Segmentation and Analysis of F/A-18

Making informed decisions regarding a vanpool's cost can be challenging. Simply comparing commuting costs is not enough. One vanpool commuting cost can be similar to another, yet it still may be overcharging the riders. To fully understand the context of the vanpool's cost, you must review and compare the components that make up the cost (such as round trip distance, gas mileage, subsidies, vehicle efficiency, etc.). Doing such a detailed review is difficult to achieve for the average user. This research aims to simplify the process.

Using machine learning to make informed decisions regarding vanpool commuting cost

Title: PRO-GAGE: A High Performance Compact GAGE Hash Function Processor for Space Applications The emergence of the Internet of Things (IoT) applications presents a wide range of opportunities in areas such as automotive, consumer, energy, smart city, and many others. Since many of these systems operate in environments with heightened sensitivity and privacy, they require a trusted security level. The implemented security measures must fit within strict power and area budgets, yet deliver maximum performance mandated by the application. The need for small devices as CubeSats, airships, and other new small devices generation, which can be close to the earth or far from it, which considers deep-space, is elevating. The security of these devices is also a fundamental matter as the number of players (both private and public sectors) in exploring space is in continuous increase. One of the critical aspects is data communication, as open space now is considered an insecure channel, which increases the need to constrain security primitives on these flying devices to keep it secure. At the same time, it will affect the battery lifetime of it. In this paper, we propose and evaluate the area, power and performance of a novel High Performance Compact low-power GAGE hash function processor (PRO-GAGE). The results show PRO-GAGE is 29x faster, and 28x more power-efficient than the reported implementation of GAGE hash function on AVR ATMEGA 328p micro-controller. PRO-GAGE needs approximately 2.6 times less power than several comparable LWC designs. In addition, the proposed processor performs approximately 2.6 times longer than the other LWC designs when powered by typical Ni-Cd battery

PRO-GAGE: A High Performance Compact GAGE Hash Function Processor for Space Applications

Information and influence operations (IO) are a tool for advancing adversarial agendas in near-peer conflicts by augmenting narratives that drive actions in communities. As evidenced by the 2016 US Presidential Election, nation-state actors employ a variety of tactics, techniques, and procedures (TTPs), such as bot nets (semi-automated) and troll farms (human-powered), to conduct IO via text-driven communications. Countering IO has become a priority for the national security community. However, the national security community currently lacks a capability that can emulate the large-scale, adversarial IO activity, limiting our ability to train appropriate responses. Here we present a case study in using GPT-2, a state-of-the-art Transformer-based generative language model, as a synthetic opposition force for real-time IO activity emulation. We show that fine-tuning GPT-2 models with specific datasets can effectively emulate different personas in an information landscape. We differentiate model personas by measuring various language dimensions – emotion, sentiment, morality – using VaderSentiment and custom BERT-based classifiers. We further evaluate the generation quality using measures of local and global coherence, sentence quality, and generation velocity. Our initial results demonstrate that generative language models will be successful in emulating various personas engaged in a coordinated IO campaign. Further, these results suggest research should be conducted to understand appropriate mechanisms for prompting (i.e., feeding communications to a generative language model to spur additional content generation) these models in a way that furthers the desired narrative. We anticipate that the effective application of fine-tuning and prompting techniques will support the use of generative language models as a synthetic opposition in large-scale training exercises. Operating these models as a synthetic adversary will spur advances in the human-AI collaboration and training fields that improve our understanding of command-and-control (i.e., white cell to red force) and action-reaction (i.e., blue force to red force) relationships in IO response training. These advances will be useful in extending this synthetic content generation capability to other domains and media.

Realistic Cacophony: Creating Synthetic Personas with Diverse Sentiment, Emotion, and Morality for Influence Operations Training

Accurate positioning and navigation are critical to a wide variety of Navy and DoD applications. A Global Positioning System (GPS) and an Inertial Navigation System (INS) make up the backbone of modern navigation systems. Even the best inertial sensors will drift over time without some sort of correction. GPS typically provides this correction, but is not always available due to environmental or adversarial factors. When this is the case, many systems and missions are limited by the inertial system’s holdover time (the length of time the system is able to provide useful position solutions without corrections). The DIRT-I project aims to increase this holdover time by applying reinforcement learning techniques to the inertial system’s position solution.

DIRT-I: Drift Improvement through Reinforement Learning - Inertial Navgation Systems

Autonomous systems have an array of specialized sensors to observe their surroundings. An autonomous platform must be able to combine many disparate streams of information in order to leverage all of the available information while creating a comprehensive model of its environment. Deep learning-based object detectors typically need to be trained on swaths of labeled data to achieve good performance. Unfortunately, collecting and labeling multimodal data in the real world is exceedingly labor-intensive. The use of video game graphics engines in the production of images and video has emerged as a relatively cheap and effective way to create new datasets. A unique aspect of using gaming engines to generate data is the ability to introduce domain randomization which randomizes certain parameters of the game engine and generation scheme in order to improve generalization to real-life data. In this paper, we outline the creation of a multi-modal synthetic dataset using domain randomization. Our dataset will focus on the two most popular sensors in autonomous vehicles, LiDAR and RGB cameras. We will discuss the effect of different parameters on each modality and perform baseline testing of an object detector using a data-fusion deep learning architecture on both our synthetic dataset and the KITTI dataset for comparison.

Multi-modal Dataset Generation using Domain Randomization for Object
Detection

We present initial research on physics-mathematics modeling, computational simulation, and data visualization that aim at discovering the physics of the adversarial minds of threat actors who develop exploits and malware. In this research, the physics of an adversarial mind refer to laws that govern the dynamics of human cognition as reflected in malware. The physics of adversarial minds, in turn, may enable us to discover cognitive vulnerabilities that are exploitable through defensive cyber deception to persistently detect malware across many generations, or for as long as the malware are developed by threat actors whose cognition we have analyzed and have found to be vulnerable. We aim at discovering physics of cognition that pinpoint innate and hence possibly involuntary cognitive vulnerabilities, as well as developmental cognitive vulnerabilities. We expect these cognitive vulnerabilities to be persistent and recurring in a variety of manifestations. The enabler of the discovery of cognitive vulnerabilities in cyber space is a deep and thorough understanding of how the cognition of threat actors and cyber terrain interact with each other to produce exploits and malware. We aim at discovering the interplay between computing and the cognitive processes that take place in the mind of threat actors. In cognitive science, human cognition is often characterized via mathematical models. Our goal is to extend and adapt these models to operations in cyber space, and thus integrate them with elements of computing. A fusion of cognitive phenomena and computing in a mathematical form serves as basis for this research, which is envisioned to leverage cyber mechanisms to learn and then probe the cognition of threat actors for cognitive vulnerabilities. A key approach in this research is the small-scale decomposition of the presence of human mind in malware into particles. We develop small-scale mathematical models that characterize the individual action of those particles, their physical quantities, and how those physical quantities are related to each other. We encode in these small-scale models the inner workings of cognition in ways that resemble physical matter such as gasses and fluids, while representing the physics of cognition with high fidelity. We then exploit small-scale models to develop large-scale mathematical models of malware-reflected cognition. We discover physical properties of an adversarial mind in aggregate from a probabilistic examination via statistical physics and nonlinear dynamical systems theory of the underlying logical particles. Of particular interest are macroscopic physical quantities of cognition that emerge from the collective behavior of particles. In other words, this research discovers the thermodynamics of an adversarial mind using malware as measurable observable. This research explores statistical physics and dynamical systems theory similarly to how these disciplines were used to study gasses and fluids in conventional physics. By developing high-fidelity mathematical models that establish equivalence between cognition in malware and physical matter, this research creates potential for inheriting the past success of those methods.

Premium content

Next from NAML 2021

Predictive Analytics in the Maritime Domain

Similar lecture

Deep Feature Space Trojan Attack of Neural Networks by Controlled Detoxification