Search NASA⌕ Search

SEARCH · Search NASA

Results for “point adversarial attack”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 19 records

Point Adversarial Self-Mining: A Simple Method for Facial Expression Recognition

In this article, we propose a simple yet effective approach, called point adversarial self mining (PASM), to improve the recognition accuracy in facial expression recognition (FER). Unlike previous works focusing on designing specific architectures or loss functions to solve this problem, PASM boosts the network capability by simulating human learning processes: providing updated learning materials and guidance from more capable teachers. Specifically, to generate new learning materials, PASM leverages a point adversarial attack method and a trained teacher network to locate the most informative position related to the target task, generating harder learning samples to refine the network. The searched position is highly adaptive since it considers both the statistical information of each sample and the teacher network capability. Other than being provided new learning materials, the student network also receives guidance from the teacher network. After the student network finishes training, the student network changes its role and acts as a teacher, generating new learning materials and providing stronger guidance to train a better student network. Here, the adaptive learning materials generation and teacher/student update can be conducted more than one time, improving the network capability iteratively. Extensive experimental results validate the efficacy of our method over the existing state of the arts for FER.

97 MATHEMATICS AND COMPUTING↗

Transferable Adversarial Attack on 3D Object Tracking in Point Cloud

3D point cloud object tracking has recently witnessed considerable progress relying on deep learning. Such progress, however, mainly focuses on improving tracking accuracy. The risk, especially considering that deep neural network is vulnerable to adversarial perturbations, of a tracker being attacked is often neglected and rarely explored. In order to attract attentions to this potential risk and facilitate the study of robustness in point cloud tracking, we introduce a novel transferable attack network (TAN) to deceive 3D point cloud tracking. Specifically, TAN consists of a 3D adversarial generator, which is trained with a carefully designed multi-fold drift (MFD) loss. The MFD loss considers three common grounds, including classification, intermediate feature and angle drifts, across different 3D point cloud tracking frameworks for perturbation generation, leading to high transferability of TAN for attack. In our extensive experiments, we demonstrate the proposed TAN is able to not only drastically degrade the victim 3D point cloud tracker, \ie, P2B, but also effectively deceive other unseen state-of-the-art approaches such as BAT and M^2Track, posing a new threat to 3D point cloud tracking.

97 MATHEMATICS AND COMPUTING↗

Nuclear power plant instrumentation and control cyber security common vector access leading to relational common cause failures - 340

Nuclear power plants and facilities have been implementing digital system upgrades into their previously analog systems for well over twenty years. New nuclear facilities' control, security, and emergency preparedness systems are almost exclusively built on digital architectures with a high degree of communication between the various systems that are often integrated together into a central control station to aid in operation or security of the facility. As digital systems become more widespread in nuclear facility control system architectures, cyber security related issues have become a significant concern to operators, regulators, governments, and other groups. Among the many concerns related to digital systems and cyber security is the area of common cause and common mode failures. This paper introduces, defines, and discusses some sources of common cause failure from a cyber security perspective: common vector access. This refers to specific access points that an adversary can exploit through a single attack sequence that have the potential to provide relational failures through common cause on multiple components, subsystems, systems, or plants. This paper will further discuss interconnected processes where these access points may exist, the importance of limiting or controlling these pinch points, and some methods of protecting common vector access points. (authors)

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Automated Membership Inference Attacks: Discovering MIA Signal Computations using LLM Agents

Membership inference attacks (MIAs), which enable adversaries to determine whether specific data points were part of a model's training dataset, have emerged as an important framework to understand, assess, and quantify the potential information leakage associated with machine learning systems. Designing effective MIAs is a challenging task that usually requires extensive manual exploration of model behaviors to identify potential vulnerabilities. In this paper, we introduce AutoMIA -- a novel framework that leverages large language model (LLM) agents to automate the design and implementation of new MIA signal computations. By utilizing LLM agents, we can systematically explore a vast space of potential attack strategies, enabling the discovery of novel strategies. Our experiments demonstrate AutoMIA can successfully discover new MIAs that are specifically tailored to user-configured target model and dataset, resulting in improvements of up to 0.18 in absolute AUC over existing MIAs. This work provides the first demonstration that LLM agents can serve as an effective and scalable paradigm for designing and implementing MIAs with SOTA performance, opening up new avenues for future exploration.

Tran, Toan Viet [Emory University]↗

Robustness of Deep Learning Classification to Adversarial Input on GPUs: Asynchronous Parallel Accumulation Is a Source of Vulnerability

The ability of machine learning (ML) classification models to resist small, targeted input perturbations—known as adversarial attacks—is a key measure of their safety and reliability. We show that floating-point non associativity (FPNA) coupled with asynchronous parallel programming on GPUs is sufficient to result in misclassification, without any perturbation to the input. Additionally, we show that this misclassification is particularly significant for inputs close to the decision boundary and that standard adversarial robustness results may be overestimated up to 4.6 when not considering machine-level details. We first study a linear classifier, before focusing on standard Graph Neural Network (GNN) architectures and datasets used in robustness assessments. We develop a novel black-box attack using Bayesian optimization to discover external workloads that can change the instruction scheduling which bias the output of reductions on GPUs and reliably lead to misclassification. Motivated by these results, we present a new learnable permutation (LP) gradient-based approach to learning floating-point operation orderings that lead to misclassifications. The LP approach provides a worst-case estimate in a computationally efficient manner, avoiding the need to run identical experiments tens of thousands of times over a potentially large set of possible GPU states or architectures. Finally, using instrumentation-based testing, we investigate parallel reduction ordering across different GPU architectures under external background workloads, when utilizing multi-GPU virtualization, and when applying power capping. Our results demonstrate that parallel reduction ordering varies significantly across architectures under the first two conditions, substantially increasing the search space required to fully test the effects of this parallel scheduler-based vulnerability. These results and the methods developed here can help to include machine-level considerations into adversarial robustness assessments, which can make a difference in safety and mission critical applications.

Shanmugavelu, Sanjif [Maxeler Technologies, a Groq↗

Discovering the Most Severe K-Point Failure Based on Reinforcement Learning: Preprint

Smart devices are essential to ensure the stability of the power grid and resilience to intermittent energy production. However, smart devices can also be the target of cyber adversaries that may exploit false data injection attacks (FDIAs) to induce unstable grid conditions. A practical consideration of FDIA mitigation approaches is addressed here: given a finite available budget, for which smart device should cyber-threat mitigation be deployed first? In this work, this question is answered by identifying the so-called most-sensitive devices, i.e., the devices that, if compromised, can let an adversary induce the most serious grid instabilities. The method proposed utilizes an adversarial reinforcement learning (RL) framework to identify the k-mostsensitive smart devices (here, smart inverters). The adversarial agent can tamper with the compromised inverters' active and reactive operating power setup points, with the goal of maximizing voltage deviations. Numerical results show that the proposed RL method finds the optimal attack scenarios for 1-point failure and the near-optimal solution for the 2-point case. Additionally, the proposed RL method achieves an 8.8 speed-up ratio in running time compared to the brute force method for the 2-point case.

97 MATHEMATICS AND COMPUTING↗

Towards A Theory of Autonomous Reconstitution of Compromised Cyber-Systems

The ability to maintain mission-critical operations in cyber-systems in the face of disruptions is critical. Faults in cyber systems can come from accidental sources (e.g., natural failure of a component) or deliberate sources (e.g., an intelligent adversary). Natural and intentional manipulation of data, computing, or coordination are the most impactful ways that an attacker can prevent an infrastructure from realizing its mission goals. Under these conditions, the ability to reconstitute critical infrastructure becomes important. Specifically, the question is: Given an intelligent adversary, how can cyber systems respond to keep critical infrastructure operational? In cyber systems, the distributed nature of the system poses serious difficulties in maintaining operations, in part due to the fact that a centralized command and control apparatus is unlikely to provide a robust framework for resilience. Resilience in cyber-systems, in general, has several components, and requires the ability to anticipate and withstand attacks or faults, as well as recover from faults and evolve the system to improve future resilience. The recovery effort (and any subsequent evolution) may require significant reconfiguration of the system (at all levels – hardware, software, services, permissions, etc.) if the system is to be made resilient to further attack or faults. This is especially important in the case of ongoing attacks, where reconfiguration decisions must be taken with care to avoid further compromising the system while maintaining continuity of operations. Collectively, we will label this recovery and evolution process as “reconstitution”. Currently, reconstitution is performed manually, generally after-the-fact, and usually consists of either standing up redundant systems, check-points (rolling back the configuration to a “clean” state), or re-creating the system using “gold-standard” copies. For enterprise systems, such reconstitution may be performed either directly on hardware, or using virtual machines. A significant challenge within this context is the ability to verify that the reconstitution is performed in a manner that renders the cyber-system resilient to ongoing and future attacks or faults. Fundamentally, the need is to determine optimal configuration of the cyber system when a fault is determined to be present. While existing theories for fault tolerance (for example, Byzantine fault tolerance) can guarantee resilience under certain conditions, in practice, these theories can break down in the face of an intelligent adversary. Further, it is difficult, in a dynamically evolving environment, to determine whether the necessary conditions for resilience have been met, resulting in difficulties in achieving resilient operation. In addition, existing theories do not sufficiently take into account the cost for attack and defense (the adversary is generally assumed to have infinite resources and time), hierarchy of importance (all network resources are assumed to be equally important), and the dynamic nature of some attacks (i.e., as the attack evolves, can resilience be maintained?). Alternative approaches to resilience based on a centralized command and control structure suffer from a single-point-failure. This paper presents preliminary research towards concepts for effective autonomous reconstitution of compromised cyber systems. We describe a mathematical framework as a first step towards a theoretical basis for autonomous reconstitution in dynamic cyber-system environments. We then propose formulating autonomous reconstitution as an optimization problem and describe some of the challenges associated with this formulation. This is followed by a brief discussion on potential solutions to these challenges.

asymmetric resilience↗

Microreactor Security-by-Design Recommendations for Domestic and International Deployments

This report outlines methods vendors can use to incorporate security-by-design (SeBD) into their microreactor facility design to support and address security for both U.S. and international deployment. The team developed a hypothetical below-grade microreactor with a physical protection system (PPS) to protect the microreactor against acts of theft and sabotage and evaluated it against two adversary attack scenarios defined by a group of adversary subject matter experts (SMEs). The hypothetical microreactor facility consists of two distinct buildings. The first is the above-grade protected area (PA) entry control point (ECP) building, which houses security personnel responsible for conducting screenings and managing access to the PA. The second building is the reactor building, which features both an above-grade floor and a below-grade floor.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Cyber security evaluation of II&C technologies

The Light Water Reactor Sustainability (LWRS) Program is a research and development program sponsored by the Department of Energy, which is conducted in close collaboration with industry to provide the technical foundations for licensing and managing the long-term, safe and economical operation of current nuclear power plants The LWRS Program serves to help the US nuclear industry adopt new technologies and engineering solutions that facilitate the continued safe operation of the plants and extension of the current operating licenses. Within the LWRS Program, the Advanced Instrumentation, Information, and Control (II&C) Systems Technologies Pathway conducts targeted research and development (R&D) to address aging and reliability concerns with the legacy instrumentation and control and related information systems of the U.S. operating light water reactor (LWR) fleet. The II&C Pathway is conducted by Idaho National Laboratory (INL). Cyber security is a common concern among nuclear utilities and other nuclear industry stakeholders regarding the digital technologies that are being developed under this program. This concern extends to the point of calling into question whether these types of technologies could ever be deployed in nuclear plants given the possibility that the information in them can be compromised and the technologies themselves can potentially be exploited to serve as attack vectors for adversaries. To this end, a cyber security evaluation has been conducted of these technologies to determine whether they constitute a threat beyond what the nuclear plants already manage within their regulatory-required cyber security programs. Specifically, the evaluation is based on NEI 08-09, which is the industry’s template for cyber security programs and evaluations, accepted by the Nuclear Regulatory Commission (NRC) as responsive to the requirements of the nuclear power plant cyber security regulation found in 10 CFR 73.54. The evaluation was conducted by a cyber security team with expertise in nuclear utility cyber security programs and experience in conducting these evaluations. The evaluation has determined that, for the most part, cyber security will not be a limiting factor in the application of these technologies to nuclear power plant applications.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Towards Provable Security in Industrial Control Systems Via Dynamic Protocol Attestation

Industrial control systems (ICSs) increasingly rely on digital technologies vulnerable to cyber attacks. Cyber attackers can infiltrate ICSs and execute malicious actions. Individually, each action seems innocuous. But taken together, they cause the system to enter an unsafe state. These attacks have resulted in dramatic consequences such as physical damage, economic loss, and environmental catastrophes. This paper introduces a methodology that restricts actions using protocols. These protocols only allow safe actions to execute. Protocols are written in a domain specific language we have embedded in an interactive theorem prover (ITP). The ITP enables formal, machine-checked proofs to ensure protocols maintain safety properties. We use dynamic attestation to ensure ICSs conform to their protocol even if an adversary compromises a component. Since protocol conformance prevents unsafe actions, the previously mentioned cyber attacks become impossible. We demonstrate the effectiveness of our methodology using an example from the Fischertechnik Industry 4.0 platform. We measure dynamic attestation's impact on latency and throughput. Our approach is a starting point for studying how to combine formal methods and protocol design to thwart attacks intended to cripple ICSs.

97 MATHEMATICS AND COMPUTING↗

Persistent Classification: Understanding Adversarial Attacks by Studying Decision Boundary Dynamics

ABSTRACT There are a number of hypotheses underlying the existence of adversarial examples for classification problems. These include the high‐dimensionality of the data, the high codimension in the ambient space of the data manifolds of interest, and that the structure of machine learning models may encourage classifiers to develop decision boundaries close to data points. This article proposes a new framework for studying adversarial examples that does not depend directly on the distance to the decision boundary. Similarly to the smoothed classifier literature, we define a (natural or adversarial) data point to be ( γ , σ)‐stable if the probability of the same classification is at least for points sampled in a Gaussian neighborhood of the point with a given standard deviation . We focus on studying the differences between persistence metrics along interpolants of natural and adversarial points. We show that adversarial examples have significantly lower persistence than natural examples for large neural networks in the context of the MNIST and ImageNet datasets. We connect this lack of persistence with decision boundary geometry by measuring angles of interpolants with respect to decision boundaries. Finally, we connect this approach with robustness by developing a manifold alignment gradient metric and demonstrating the increase in robustness that can be achieved when training with the addition of this metric.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Modeling design and control problems involving neural network surrogates

Here, we consider nonlinear optimization problems that involve surrogate models represented by neural networks. We demonstrate first how to directly embed neural network evaluation into optimization models, highlight a difficulty with this approach that can prevent convergence, and then characterize stationarity of such models. We then present two alternative formulations of these problems in the specific case of feedforward neural networks with ReLU activation: as a mixed-integer optimization problem and as a mathematical program with complementarity constraints. For the latter formulation we prove that stationarity at a point for this problem corresponds to stationarity of the embedded formulation. Each of these formulations may be solved with state-of-the-art optimization methods, and we show how to obtain good initial feasible solutions for these methods. We compare our formulations on three practical applications arising in the design and control of combustion engines, in the generation of adversarial attacks on classifier networks, and in the determination of optimal flows in an oil well network.

97 MATHEMATICS AND COMPUTING↗

Clean Energy Cybersecurity Accelerator Cohort 1: Authentication and Authorization

In the 2023 National Cybersecurity Strategy, the Biden-Harris Administration defines the need for a "defensible, resilient digital ecosystem where it is costlier to attack systems than defend them." The strategy cites the Clean Energy Cybersecurity Accelerator (CECA) as an exemplary effort to bolster the security and resilience of clean energy generation. These efforts help "secure the clean energy grid of the future and [generate] security best practices that extend to other critical infrastructure sectors" and promise broad and far-reaching impacts to bridge the capabilities of private industry and the needs of energy production. Cohort 1 of CECA launched in the fall of 2022 with a focus on solutions that provide strong authentication and authorization for industrial control systems to mitigate attacks on the energy grid. Authentication and authorization verify that the identity (authentication) and permissions (authorization) of a user or device are aligned with their assigned roles. Weaknesses in either can have serious repercussions. To assess the strength of Cohort 1's solutions, CECA devised threat scenarios grounded in historical precedents: the CECA team reviewed exploits from real-world case studies of state-sponsored actors to match the assessment's attack paths and targets. Cohort 1 results provided the energy industry, product vendors, and related agencies valuable insights into the efficacy and applicability of solutions in common system configurations under realistic threat scenarios. The results of the assessment highlight points for interrogation and improvement in subsequent technology iterations. CECA's evaluations are part of an ongoing conversation and collaboration to bolster U.S. cyber resilience against adversaries today and in the future.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Proactive routing mutation against stealthy Distributed Denial of Service attacks: metrics, modeling, and analysis

The Infrastructure Distributed Denial of Service (IDDoS) attacks continue to be one of the most devastating challenges facing cyber systems. The new generation of IDDoS attacks exploit the inherent weakness of cyber infrastructure including deterministic nature of routes, skew distribution of flows, and Internet ossification to discover the network critical links and launch highly stealthy flooding attacks that are not observable at the victim end. In this paper, first, we propose a new metric to quantitatively measure the potential susceptibility of any arbitrary target server or domain to stealthy IDDoS attacks, and es- timate the impact of such susceptibility on enterprises. Second, we develop a proactive route mutation technique to minimize the susceptibility to these attacks by dynamically changing the flow paths periodically to invalidate the adversary knowledge about the network and avoid targeted critical links. Our proposed approach actively changes these network paths while satisfying security and qualify of service requirements. We present an integrated approach of proactive route mutation that combines both infrastructure-based mutation that is based on reconfiguration of switches and routers, and middle-box approach that uses an overlay of end-point proxies to construct a virtual network path free of critical links to reach a destination. We implemented the proactive path mutation technique on a Software Defined Network using the OpendDaylight controller to demonstrate a feasible deployment of this approach. Our evaluation validates the correctness, effectiveness, and scalability of the proposed approaches.

ARC↗

Complete Evaluation on Advanced Reactor Machine Learning Subversion Attacks (Final)

Navigating through the world of Artificial Intelligence (AI) in nuclear reactors and their Instrumentation and Control (I&C) systems demands a careful, deliberate journey. AI’s capability to manage massive datasets and streamline control systems has indeed carved out a significant role in various sectors, including nuclear energy. However, while AI, and particularly Large Language Models (LLMs), bring a lot to the table in terms of operational efficiency and anomaly detection, they also expose the sector to a new breed of cybersecurity threats, like Inference Attacks, Adversarial Attacks, and Trojan Attacks. This guide is designed to be a straightforward manual, diving deep into the intertwining worlds of AI and cybersecurity within nuclear reactors, and tailoring insights for three crucial audiences: I&C Vendors/Developers, Nuclear Regulators, and Nuclear Reactor Operators and Cyber Defense Teams. (1) Section 2, directed at I&C Vendors/Developers, will provide a clear and focused look at several cybersecurity attacks, offering practical recommendations and detailed scenarios related to AI cybersecurity. This section isn’t just about identifying problems but also about giving solid, usable solutions. (2) Section 3, meant for Nuclear Regulators, gets straight to the point about regulations, policy suggestions, and guidelines that are needed to lay down a robust, secure, and ethical foundation for the application of AI in nuclear operations. The focus is on making sure that everything adheres to international standards and laws while being practicable and clear-cut. (3) Section 4, aimed at Nuclear Reactor Operators and Cyber Defense Teams, offers an exhaustive exploration and technical reports, with clear recommendations and scenario analyses vital to protect operational environments and guarantee the secure application of AI in nuclear reactor operations. The goal is simple: as we step into an era where AI becomes a fundamental element of our technological and energy infrastructures, this guide is here to act as a clear, direct handbook, ensuring that AI is implemented within the nuclear sector in a manner that is secure, responsible, and practical. It’s about striking a balance – optimizing the undeniable benefits offered by AI while securing and shielding against potential cyber threats as we move through this new and complex landscape.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Employing a Hardware-in-the-Loop Approach to Realize a Fully Homomorphic Controller for a Small Modular Advanced High Temperature Reactor

This paper addresses the cybersecurity challenges of advanced nuclear reactors by integrating fully homomorphic encryption (FHE) into their control systems, enabling encrypted processing of control signals without compromising functionality. Advanced nuclear reactors, including Small Modular Reactors (SMRs) and microreactors, aim to achieve autonomous and remote operations, reducing costs and enhancing competitiveness. However, these advancements expand the attack surface for cyberattacks, particularly in autonomous and remote operation scenarios. Cyberattacks can exploit vulnerabilities to manipulate physical processes, causing shutdowns, asset damage, or public harm. Such attacks begin with passive reconnaissance, where adversaries intercept communications or observe behaviors to gather information, which is then leveraged to execute cyber-physical attacks by injecting malicious commands. Nuclear power must adopt cybersecurity protection measures to secure the integrity and availability of their digital control systems. This paper demonstrates the application of FHE to secure operations by enabling encrypted processing of sensitive signals and parameters -- ensuring privacy without exposing data. FHE supports secure mathematical operations on encrypted data without requiring decryption. Using a hardware-in-the-loop (HIL) approach, this paper implements an FHE-integrated controller on a BeagleBone Black (BBB) controlling a simulation of the Small Modular Advanced High Temperature Reactor (SmAHTR). By doing so, the encrypted controller protects the integrity of critical set points and control signals during transmission and processing. Thus, FHE-integrated controllers enhance secure operations of advanced nuclear reactors while maintaining functionality.

control systems↗

A Moving Target Defense Control Framework for Cyber-Physical Systems

This paper considers the problem of efficiently and securely controlling cyber-physical systems that are operating in uncertain, and adversarial environments. To mitigate sensor, actuator attacks, and performance loss due to such attacks, we formulate a secure control algorithm that consists of a proactive and a reactive defense mechanism. The proactive mechanism, which is based on the principles of moving target defense, utilizes a stochastic switching structure to dynamically and continuously alter the parameters of the system, while hindering the attacker's ability to conduct successful reconnaissance to the system. The unpredictability of the current actuator and sensor is optimized using an information entropy measure, which is induced by probabilistic switching. The reactive mechanism on the other side, detects potentially attacked components, namely sensors and actuators, by leveraging online data to compute an integral Bellman error. A rigorous mathematical framework is presented to guarantee the stability of the equilibrium point of the closed-loop system, and provide a quantified bound on the performance loss when utilizing both reactive and proactive mechanisms. Simulation results show the efficacy of the proposed approaches on a benchmark aircraft model.

42 ENGINEERING↗

AntBot: Anti-pollution peer-to-peer botnets

Botnets, which are responsible for many email sparnming and DDoS (Distributed Denial of Service) attacks in the current Internet, have emerged as one of most severe cyber-threats in recent years. To evade detection and improve resistance against countermeasures, botnets have evolved from the first generation that relies on IRC chat channels to deliver commands to the current generation that uses highly resilient P2P (Peer-to-Peer) protocols to spread their C&C (Command and Control) information. It is, however, revealed that P2P botnets, although relieved from the single point of failure that IRC botnets suffer, can be easily disrupted using pollution-based mitigation schemes [15]. In this paper, we play the devil's advocate and propose a new type of hypothetical botnets called AntBot, which aim to propagate their C&C information to individual bots even though there exists an adversary that persistently pollutes keys used by seized bots to search the command information. The key idea of AntBot is a tree-like structure that bots use to deliver the command so that captured bots reveal only limited information. To evaluate effectiveness of AntBot against pollution-based mitigation in a virtual environment, we develop a distributed P2P botnet simulator. Using extensive experiments, we demonstrate that AntBot operates resiliently against pollution-based mitigation. We further present a few potential defense schemes that could effectively disrupt AntBot operations.

97 MATHEMATICS AND COMPUTING↗