Millimeter wave attacks with meta-material-enhanced attack tags

WO2024178210A3PCT designated stage expired Publication Date: 2024-11-07THE REGENTS OF THE UNIVERSITY OF COLORADO
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
PCT/US2024/016873
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-02-22
Filing Date
2024-02-22
Publication Date
2024-11-07

AI Technical Summary

Technical Problem

Millimeter wave (mmWave) sensing systems are vulnerable to attacks that can lead to severe consequences, such as vehicle crashes or malicious intrusions, due to their susceptibility to active electronic-based approaches which are expensive and easily detectable, while existing passive attack methods are not practical for widespread use.

Method used

The development of meta-material-enhanced passive mmWave attack tags, specifically designed to disrupt RF signals by absorbing, reflecting, or modifying the polarization of mmWave signals, using materials like C-RAM LF, tin foil, or copper parallel wire grids, optimized through a simulator-based framework to achieve high attack accuracy and cost-effectiveness.

Benefits of technology

The meta-material-enhanced passive mmWave attack tags achieve up to 97% accuracy in range estimation, 96% in angle estimation, and 91% in speed estimation, while being significantly cheaper than existing methods, demonstrating a practical and pervasive solution for mmWave sensing attacks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure US2024016873_07112024_PF_FP_ABST
    Figure US2024016873_07112024_PF_FP_ABST
Patent Text Reader

Abstract

Systems and methods for generating and optimizing passive millimeter-wave (mmWave) attack tags are disclosed. In some implementations, optimizing a passive mmWave attack tag includes receiving parameters for a target and a scene of a mmWave attack, generating a set of initial parameters for a passive mmWave attack tag, determining an optimized set of parameters for the passive mmWave attack tag by iteratively modifying the set of initial parameters for the passive mmWave attack tag until a calculated attack loss is less than a predefined threshold or a maximum number of iterations is reached, and outputting the optimized set of parameters for the passive mmWave attack tag.
Need to check novelty before this filing date? Find Prior Art

Description

MILLIMETER WAVE ATTACKS WITH META-MATERIAL-ENHANCEDATTACK TAGSCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to and the benefit of U.S. Provisional Patent Application No. 63 / 486,281, filed February 22, 2023, which is incorporated herein by reference in its entirety.BACKGROUND

[0002] Millimeter wave (mmWave) refers to a band of radio frequencies that operate in the high-frequency portion of the electromagnetic spectrum. Specifically, mmWave frequencies range from 30 to 300 gigahertz (GHz), corresponding to wavelengths between 1 millimeter and 10 millimeters. The characteristics of mmWave frequencies make them suitable for various applications, including telecommunications, radar systems, imaging, and more. In this regard, mmWave sensing has been employed extensively in many different use cases (e.g., advanced driving aids for vehicles, security systems, robotics, etc.) due to its exceptional ability to determine a target’s range, angle, and velocity. For example, mmWave sensing is known to operate in harsh weather and climate conditions (e.g., lighting, smoke, and fog). mmWave radars have been used for obstacle detection, speed measurement, localization, invasion detection, blind-spot detection, collision avoidance, and surveillance.

[0003] Due to its widespread use, interest has developed in investigating different techniques and avenues for attacks on mmWave sensing systems and devices, e.g., to identify solutions to mitigate attacks. For example, a mmWave attack could be employed to interfere with measurement sensing, for example, to cause a normally sensed object (e.g., a vehicle, a road obstacle) to disappear, which can result in serious consequences such as a vehicle crash in an automotive application or malicious intrusion in security applications. As another example, mmWave attacks can cause security systems, robots, and vehicles to experience a ghosting effect, whereby a mmWave sensing radar detects an object that does not exist (e.g., resulting in a false alarm, phantom braking, etc.). Such erroneous sensing mechanisms can further be designed for meticulously targeted attacks (e.g., attacks on mmWave sensing). Thus, mmWave sensing units are prone to sensing errors that can lead to severe consequences.SUMMARY

[0004] One implementation of the present disclosure is a system including: a processor; and memory having instructions stored thereon that, when executed by the processor, cause the system to: receive target parameters and scene parameters for a potential millimeter-wave (mmWave) attack on a target entity; generate a set of initial parameters for a passive mmWave attack tag, wherein the mmWave attack tag is configured to disrupt radiofrequency (RF) signals emitted by the target entity; automatically determine an optimized set of parameters for the passive mmWave attack tag by iteratively modifying the set of initial parameters for the passive mmWave attack tag until a calculated attack loss is less than a predefined threshold and / or a maximum number of iterations is reached; and output, to a user of the system, the optimized set of parameters for the passive mmWave attack tag, wherein the passive mmWave attack tag is used to conduct tests with respect to the target entity under real-world conditions.

[0005] In some implementations, the passive mmWave attack tag is at least partially formed from one or more of Counter Rocket, Artillery, and Mortar - Lightweight System (C-RAM LF), tin foil, or a copper parallel wire grid.

[0006] In some implementations, the passive mmWave attack tag is a malleable two- dimensional or three-dimensional structure.

[0007] In some implementations, the passive mmWave attack tag includes at least one of a hexagonal shape or honeycomb structure.

[0008] In some implementations, the passive mmWave attack tag is configured to adhere to a surface of at least one of a vehicle, autonomous robot, or stationary structure during the tests.

[0009] In some implementations, the passive mmWave attack tag is configured to disrupt the RF signals emitted by the target entity by one of absorbing at least a portion of the RF signals, reflecting a majority of the RF signals back to the target entity, or modifying a polarization of the RF signals.

[0010] In some implementations, the target parameters include at least one of a frequency, bandwidth, polarization mode, or sensing distance of a radar sensor operated by the target entity.

[0011] In some implementations, the scene parameters include a geometry and / or material properties of at least one object within a mmWave attack environment.

[0012] In some implementations, the set of initial parameters includes at least one of a material, a pattern, a size, a position, and a rotation of the passive mmWave attack tag.

[0013] In some implementations, the RF signals emitted by the target entity are between 24 GHz and 300 GHz.

[0014] Another implementation of the present disclosure is a method including: receiving target parameters and scene parameters for a potential millimeter-wave (mmWave) attack on a target entity; generating a set of initial parameters for a passive mmWave attack tag, wherein the passive mmWave attack tag is configured to disrupt radiofrequency (RF) signals emitted by the target entity; automatically determining an optimized set of parameters for the passive mmWave attack tag by iteratively modifying the set of initial parameters for the passive mmWave attack tag until a calculated attack loss is less than a predefined threshold and / or a maximum number of iterations is reached; and outputting, to a user, the optimized set of parameters for the passive mmWave attack tag, wherein the passive mmWave attack tag is used to conduct tests with respect to the target entity under real-world conditions.

[0015] In some implementations, the passive mmWave attack tag is at least partially formed from one or more of C-RAM LF, tin foil, or a copper parallel wire grid.

[0016] In some implementations, the passive mmWave attack tag includes a malleable two- dimensional or three-dimensional structure.

[0017] In some implementations, the passive mmWave attack tag is configured to disrupting the RF signals emitted by the target entity by one of absorbing at least a portion of the RF signals, reflecting a majority of the RF signals back to the target entity, or modifying a polarization of the RF signals.

[0018] In some implementations, the target parameters include at least one of a frequency, bandwidth, polarization mode, or sensing distance of a radar sensor operated by the target entity.

[0019] In some implementations, the scene parameters include a geometry and material properties of at least one object within a mmWave attack environment.

[0020] In some implementations, the set of initial parameters includes at least one of a material, a pattern, a size, a position, and a rotation of the passive mmWave attack tag.

[0021] In some implementations, the RF signals emitted by the target entity are between 24 GHz and 300 GHz.

[0022] Yet another implementation of the present disclosure is a method including: obtaining target parameters and scene parameters for a potential millimeter-wave (mmWave) attack on a target entity; obtaining parameters for a passive mmWave attack tag, wherein the passive mmWave attack tag is configured to disrupt radiofrequency (RF) signals emitted by the targetentity; generating a digital attack environment for the mmWave attack based on the target parameters and the scene parameters; simulating emission of the RF signals emitted by the target entity within the digital attack environment; calculating reflected RF signals from the passive mmWave attack tag based on the parameters of the passive mmWave attack tag; determining simulated time-domain RF signals from the reflected RF signals; and using the simulated timedomain RF signals to optimize the parameters of the passive mmWave attack tag.

[0023] In some implementations, the target parameters include at least one of a frequency, bandwidth, polarization mode, or sensing distance of a radar sensor operated by the target entity.

[0024] In some implementations, the scene parameters include a geometry and material properties of at least one object within a mmWave attack environment.

[0025] In some implementations, the reflected RF signals are calculated using ray tracing.

[0026] Additional advantages will be set forth in part in the description which follows or may be learned by practice. The advantages will be realized and attained by means of the elements and combinations particularly pointed out in the appended claims. It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive, as claimed.BRIEF DESCRIPTION OF THE DRAWINGS

[0027] FIG. l is a block diagram of a system for determining design parameters for a passive millimeter-wave (mmWave) attack tag, according to some implementations.

[0028] FIG. 2 is a flow chart of a process for optimizing the parameters of a passive mmWave attack tag, according to some implementations.

[0029] FIG. 3 is a flow chart of a process for simulating a mmWave attack using a passive mmWave attack tag, according to some implementations.

[0030] FIG. 4A is a schematic diagram showing an example MetaWave framework.

[0031] FIG. 4B, FIG. 4C, FIG. 4D, and FIG. 4E are schematic diagrams illustrating different MetaWave attack scenarios.

[0032] FIG. 4F is a schematic diagram showing an example MetaWave system.

[0033] FIG. 4G is a schematic diagram depicting targeted absorption, reflection, and polarization of incident mmWave signals.

[0034] FIG. 4H and FIG. 41 are graphs depicting ranged amplitude analysis using Fast Fourier Transform.

[0035] FIG. 4J shows example two-dimensional (2D) and three-dimensional (3D) MetaWave tags.

[0036] FIG. 4K is a schematic diagram depicting an example MetaWave attack framework.

[0037] FIG. 4L is a schematic diagram depicting an example MetaWave simulator.

[0038] FIG. 5A shows an example system implementation including metamaterial tags.

[0039] FIG. 5B is a table showing results of a comparison between the real-world echo signal and the designed RF simulator.

[0040] FIG. 5C shows experimental results from simulated attacks.

[0041] FIG. 6 is a series of images showing various environments that were utilized in some of the conducted studies.

[0042] FIG. 7A is a graph showing overall performance of attacking range, angle, and speed sensing measurements.

[0043] FIG. 7B shows experimental results demonstrating the impact of the MetaWave Tag Optimizer on attack performance on range and angle measurements.

[0044] FIG. 7C is a graph showing performance under different environmental dynamics, including temperature, humidity, lighting, and magnetism respectively.

[0045] FIG. 7D is a graph showing the longitude evaluation of a span of 12 months.

[0046] FIG. 7E is a graph showing performance under different tag sizes.

[0047] FIG. 7F, FIG. 7G, and FIG. 7H show results from evaluation of attacks on the range, angle, and speed measurements, respectively.

[0048] FIG. 71 and FIG. 7J show experimental results from various scenarios described herein.

[0049] FIG. 8 A shows an experimental setup for a multi-object attack.

[0050] FIG. 8B and FIG. 8C are range Fast Fourier transform (FFT) heatmaps.

[0051] FIG. 8D and FIG. 8E show experimental attack results.

[0052] FIG. 8F shows operational examples depicting disguised MetaWave tags.

[0053] Various objects, aspects, features, and advantages of the disclosure will become more apparent and better understood by referring to the detailed description taken in conjunction withthe accompanying drawings, in which like reference characters identify corresponding elements throughout. In the drawings, like reference numbers generally indicate identical, functionally similar, and / or structurally similar elements.DETAILED DESCRIPTION

[0054] As mentioned above, mmWave sensing attacks (also referred to herein as “mmWave attacks”) are important to understand so that mitigation techniques can be developed. Given the unique properties of mmWave signals (recognized from about 24GHz-300GHz), current attack vectors on mmWave devices generally rely on active electronic-based approaches (e.g., active jamming of electromagnetic waves and injecting malicious signals). However, such attack signals are generated from high-cost, and proprietary equipment, meaning that active or electronic-based attacks are expensive, easily detectable, and preventable.

[0055] In contrast, non-electronic or “passive” meta-material tags are contemplated herein, which can passively modulate radiofrequency (RF) signals (e.g., in the mmWave range of 24GHz-300GHz) for communication and networking applications. Meta-material is a new class of functional materials designed around unique patterns or structures, which cause them to interact with RF signals and other forms of energy in ways not found in nature. Given these tag materials’ low cost and ubiquity, a meta-material tag can be custom designed to tamper the modulation of mmWave signals, thereby allowing for more practical and pervasive attacks on a range of mmWave sensing applications. Herein, passive mmWave attack tags, along with a system and method for optimizing attack tag parameters, are described. Together, the disclosed passive mmWave attack tags and design optimization system and methods are referred to as “MetaWave”.

[0056] Referring generally to the figures, systems and methods for designing and optimizing passive mmWave attack tags are shown, according to various implementations. At a high level, a passive mmWave attack tag is a physical device that implements a mmWave attack by disrupting RF signals emitted by the target. The passive mmWave attack tags contemplated herein are generally at least partially formed of a meta-material such as Counter Rocket, Artillery, and Mortar - Lightweight System (C-RAM LF), tin foil, or a copper parallel wire grid. In some implementations, the passive mmWave attack tags are malleable. Generally, a passive mmWave attack tag disrupts the RF signals emitted by the target by either absorbing at least a portion of the RF signals, reflecting a majority of the RF signals back to the target, or modifying a polarization of the RF signals.

[0057] Embodiments of the present disclosure provide systems and methods relating to a passive mmWave attack, referred to herein as MetaWave, with low-cost and easily obtainable meta-material tags for both vanish and ghost attack types. These meta-material tags can be made from commercial off-the-shelf (COTS) materials with customized tag designs to attack various goals, which considerably lower the attack bar on mmWave sensing. Specifically, tags made of ordinal material (e.g., C-RAM LF, tin foil, and / or a copper parallel wire grid) can be leveraged to precisely tamper with the mmWave echo signal and spoof the range, angle, and speed sensing measurements. In some implementations, an example MetaWave tag is a malleable 2D or 3D structure, such as a hexagon and / or honeycomb structure. To optimize the attack, a general simulator-based MetaWave attack framework is proposed and designed to simulate the tag modulation effects on the mmWave signal with advanced tag and scene parameters. In some implementations, the passive mmWave attack tag (“MetaWave tag”) is configured to adhere to a surface of a vehicle, autonomous robot, or stationary structure during testing. Embodiments of the present disclosure include MetaWave, a meta-material tag attack used in both simulation and real-world experiments (i.e., 20 different environments) with various attack settings.

[0058] The MetaWave tags can be used to test mmWave and / or imaging systems that are used in various applications including autonomous and semi -autonomous vehicles, robots, drones, and the like, for example, prior to manufacturing and / or deployment of such systems. Experimental results demonstrate that MetaWave can achieve up to 97% Top-1 attack accuracy on range estimation, 96% on angle estimation, and 91% on speed estimation in actual practice. Additionally, the developed MetaWave is 10-100X cheaper than existing mmWave attack methods. The usability and robustness of MetaWave were evaluated under different real -world scenarios. Moreover, in-depth analysis and discussion are provided on countermeasures for MetaWave mmWave attacks to improve wireless sensing and cyber-infrastructure security.

[0059] Referring now to FIG. 1, a block diagram of a system 100 for determining design parameters (e.g., optimal parameters) for a passive mmWave attack tag is shown, according to some implementations. At a high level, system 100 is shown to receive target parameters, scene parameters, and optionally initial passive mmWave attack tag parameters. System 100 uses these input parameters to simulate performance of a passive mmWave attack tag in a digital representation of an attack environment and, based on the simulation, can fine-tune parameters of the passive mmWave attack tag. In some implementations, system 100 outputs or provides (e.g., to a user, via a display, or the like) optimized parameters for the passive mmWave attack tag and / or deployment parameters.

[0060] System 100, also referred to herein as the “MetaWave” attack framework, is shown to include a processing circuit 102 that includes a processor 104 and a memory 410. Processor 104 can be a general-purpose processor, an ASIC, one or more FPGAs, a group of processing components, or other suitable electronic processing structures. In some implementations, processor 104 is configured to execute program code stored on memory 110 to cause system 100 to perform one or more operations, as described below in greater detail. It will be appreciated that, in implementations, system 100 is part of another computing device (e.g., a general- purpose computer) such that the components of system 100 described herein are shared with, or the same as, the host device.

[0061] Memory 110 can include one or more devices (e.g., memory units, memory devices, storage devices, etc.) for storing data and / or computer code for completing and / or facilitating the various processes described in the present disclosure. In some implementations, memory 110 includes tangible (e.g., non-transitory), computer-readable media that stores code or instructions executable by processor 104. Tangible, computer-readable media refers to any physical media that is capable of providing data that causes system 100 to operate in a particular fashion. Example tangible, computer-readable media may include, but is not limited to, volatile media, non-volatile media, removable media and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Accordingly, memory 110 can include RAM, ROM, hard drive storage, temporary storage, non-volatile memory, flash memory, optical memory, or any other suitable memory for storing software objects and / or computer instructions. Memory 110 can include database components, object code components, script components, or any other type of information structure for supporting the various activities and information structures described in the present disclosure. Memory 110 can be communicably connected to processor 104, such as via processing circuit 102, and can include computer code for executing (e.g., by processor 104) one or more processes described herein.

[0062] While shown as individual components, it will be appreciated that processor 104 and / or memory 110 can be implemented using a variety of different types and quantities of processors and memory. For example, processor 104 may represent a single processing device or multiple processing devices. Similarly, memory 110 may represent a single memory device or multiple memory devices. Additionally, in some implementations, system 100 may be implemented within a single computing device (e.g., one server, one housing, etc.). In other implementations, system 100 may be distributed across multiple servers or computers (e.g., that can exist in distributed locations). For example, system 100 may include multiple distributed computingdevices (e.g., multiple processors and / or memory devices) in communication with each other that collaborate to perform operations. For example, but not by way of limitation, an application may be partitioned in such a way as to permit concurrent and / or parallel processing of the instructions of the application. Alternatively, the data processed by the application may be partitioned in such a way as to permit concurrent and / or parallel processing of different portions of a data set by the two or more computers. For example, virtualization software may be employed by system 100 to provide the functionality of a number of servers that is not directly bound to the number of computers in system 100.

[0063] Memory 110 is shown to include an RF signal simulator 112 generally configured to simulate the emission and sensing of mmWave RF signals by a target device. A “target” device or target entity, also referred to herein as the “victim”, is the device or object that is being subjected to a mmWave attack. For example, the target device may be a vehicle or, more specifically, a mmWave radar coupled to the vehicle (e.g., for automated cruise control or obstacle detection). At a high level, RF signal simulator 112 takes scene parameters as an input and creates a digital copy of the attack trap (e.g., attack environment) based on the parameters of environmental, objects, and radar specifications. A digital radar then emits an RF signal, receives “reflected” RF signals, and approximates a number of ray samples. The ray samples are then used to calculate a ray exposure point and the reflected / refracted rays are calculated based on the surface properties of simulated passive mmWave attack tag. Radar cross-section (RCS) values are then used to calculate an echo signal (e.g., for Frequency-Modulated Continuous-Wave (FMCW)). In some implementations, RF signal simulator 112 outputs a time-domain RF signal.

[0064] In some implementations, RF signal simulator 112 is configured to determine a location where RF signals emitted by the target device intersect. In some such implementations, RF signal simulator 112 utilizes ray tracing for these calculations. The geometry of an object in a digital world is usually represented by thousands to millions of triangles (i.e., mesh). Since direct intersection computation of mesh is challenging, intersection detection with mesh is split into intersection detection with each triangle within its mesh and the integral of all intersected rays is calculated to get approximate solutions. This process is called ray tracing.

[0065] In some implementations, RF signal simulator 112 utilizes a Bounding Volume Hierarchy (BVH) structure to store the triangles, which is an optimization structure-based binary search tree that has the logarithmic time complexity of finding the nearest triangle at a given location. The Mdller-Trumbore intersection algorithm is used for the fast ray-triangle intersection calculation. The Monte Carlo method is used to calculate the approximation of all rays emitted and received by the radar. However, the classic random sampler in ray tracing andshooting and bouncing rays (SBR) is considered indifferentiable. Therefore, a new edge sampler is applied in MetaWave’s sampling state, which is continuous and can be differentiated.

[0066] Material simulation and definition are also important parameters for system 100 to consider because of the need to simulate MetaWave tags with different RF modulation methods. However, the native SBR method does not support the simulation of materials. Therefore, SBR is enhanced by introducing an extension of the Bidirectional Scattering Distribution Function (BSDF) in the ray-tracing stage. BSDF is a universal function that defines an arbitrary surface’s reflection and refraction properties. The BSDF takes incident ray and reflection ray direction as input and returns the probability of the reflection happening. In addition, it can also change the characteristics.

[0067] In some implementations, RF signal simulator 112 further estimates a RCS of simulated objects via Physical Optics Integral (POI). The POI is used to simulate the interference, diffraction, and polarization of EM wave propagation of the rays (e.g., polarization). It is worth mentioning that RF signal simulator 112 can be configured for various signal modulation technologies (e.g., FMCW, CW) in signal propagation. For example, RF signal simulator 112 can generate an FMCW waveform from a digital radar and then manipulate the transmitted signal according to the response of the objects.

[0068] Memory 110 is also shown to include a tag optimizer 114 generally configured to generate and / or fine-tune parameters of a passive mmWave attack tag. Specifically, tag optimizer 114 may operate in conjunction with RF signal simulator 112 to simulate performance of a passive mmWave attack tag and can modify parameters of the passive mmWave attack tag until the target’s simulated mmWave sensing device exhibits an error (e.g., detects a ghost object, misses a object in the detection zone, etc.). Different from most adversarial attacks that focus on the computing module (e.g., the target recognition / classification algorithms), tag optimizer 114 utilizes an iterative approach to find the optimal tag design and deployment parameters. At a high level, attack scenario parameter - which include scene parameters and / or target parameters - are received. Tag optimizer 114 then generates a digital passive mmWave attack tag in the simulator with randomly initialized tag parameters. The parameters of the digital passive mmWave attack tag are then fine-tuned and optimized through iteratively simulating the tag’s performance in the digital attack environment. Tag parameters are determined to achieve a desirable attack performance and robustness. In some implementations, the digital passive mmWave attack tag parameters are tuned until an attacking loss is less than a predefined threshold or a maximum iterations count is reached. Then, a passive mmWave attack tag designand deployment parameters can be provided such that a physical passive mmWave attack tag can be manufactured and deployed.

[0069] System 100 is also shown to include a communications interface 120 that facilitates communications between system 100 and any external components or devices. For example, communications interface 120 can provide means for transmitting data to, or receiving data from, a remote computing device (e.g., a workstation, a computer, a smartphone, a server, etc.). Accordingly, communications interface 120 can be or can include a wired or wireless communications interface (e.g., jacks, antennas, transmitters, receivers, transceivers, wire terminals, etc.) for conducting data communications, or a combination of wired and wireless communication interfaces. In some implementations, communications via communications interface 120 are direct (e.g., local wired or wireless communications) or via a network (e.g., a WAN, the Internet, a cellular network, etc.). For example, communications interface 120 may include one or more Ethernet ports for communicably coupling system 100 to a network (e.g., the Internet). In another example, communications interface 120 can include a WiFi transceiver for communicating via a wireless communications network. In yet another example, communications interface 120 may include cellular or mobile phone communications transceivers.

[0070] Referring now to FIG. 2, a flow chart of a process 200 for optimizing the parameters of a passive mmWave attack tag is shown, according to some implementations. In some implementations, process 200 is implemented by system 100, as described above. It will be appreciated that certain steps of process 200 may be optional and, in some implementations, process 200 may be implemented using less than all of the steps. It will also be appreciated that the order of steps shown in FIG. 2 is not intended to be limiting. Certain steps of process 200 may be further or more readily understood with reference to the experimental results detailed herein.

[0071] At step 202, the process includes receiving target and scene parameters for a simulated mmWave attack. In some implementations, the target and scene parameters are received from a user input (e.g., image data such as one or more images of a specific environment). In other implementations, at least some of the target and scene parameters are randomly generated. In other implementations, the target and scene parameters are retrieved from a database or are otherwise obtained by a suitable method. Target parameters, as described herein, are generally parameters and / or characteristics of a target of a mmWave attack (e.g., a vehicle’s radar system, etc.). In some implementations, target parameters include one or more of a frequency, bandwidth, polarization mode, or sensing distance of a radar sensor operated by the target. Insome implementations, scene include a geometry and material properties of any objects within a mmWave attack environment.

[0072] At step 204, the process 200 includes generating a digital copy of (e.g., simulated) a passive mmWave attack tag from a set of randomly initialized parameters. Put another way, a set of initial parameters for a passive mmWave attack tag is randomly generated as a starting point for generating a design of the passive mmWave attack tag.

[0073] At step 206, the process 200 includes generating an optimized set of parameters for the passive mmWave attack tag by iteratively modifying the initial set of parameters to minimize loss or error. In some implementations, the digital copy of the passive mmWave attack tag is used in a digital representation of an environment and the performance of the passive mmWave attack tag is estimated. Then, the parameters of the passive mmWave attack tag are adjusted (e.g., between each iteration) until a calculated attack loss is less than a threshold or until a maximum number of iterations is reached. As described herein, parameters for the passive mmWave attack tag generally include one or more of a material, a pattern, a size, a position, and a rotation of the passive mmWave attack tag.

[0074] Then, at step 208, the process includes outputting and / or presenting the optimized set of tag parameters to a user. Optionally, in some implementations, deployment parameters are also provided. Deployment parameters can include one or more of a relative size, position, and rotation of the tag. In some implementations, the process 200 can be applied to simultaneously determine optimized passive mmWave attack tag parameters for a plurality of targets (e.g., multiple radars, image sensors, combinations thereof in a plurality of modalities). Said differently, an optimized tag and / or tag parameters can be configured to simultaneously test multiple device components (e.g., multiple sensors, radars, or the like).

[0075] Referring now to FIG. 3, a flow chart of a process 300 for simulating a mmWave attack using a passive mmWave attack tag is shown, according to some implementations. In some implementations, process 300 is implemented by system 100, as described above. It will be appreciated that certain steps of process 300 may be optional and, in some implementations, process 300 may be implemented using less than all of the steps. It will also be appreciated that the order of steps shown in FIG. 3 is not intended to be limiting. Certain steps of process 300 may be further or more readily understood with reference to the experimental results detailed herein.

[0076] At step 302, the process 300 includes obtaining target and scene parameters for a mmWave attack. In some implementations, initial parameters for a passive mmWave attack tagto be tested are also received. In some implementations, the passive mmWave attack tag is configured to disrupt the RF signals emitted by a target entity (e.g., victim) by one of absorbing at least a portion of the RF signals, reflecting a majority of the RF signals back to the target entity, or modifying a polarization of the RF signals. In some examples, the target parameters comprise at least one of a frequency, bandwidth, polarization mode, or sensing distance of a radar sensor operated by the target entity. In some implementations, the scene parameters comprise a geometry and / or material properties of at least one object within a mmWave attack environment. The initial parameters can comprise at least one of a material, a pattern, a size, a position, and a rotation of the passive mmWave attack tag.

[0077] At step 304, the process 300 includes generating a digital copy or “simulated” representation of a mmWave attack environment based on the target and scene parameters.

[0078] At step 306, the process 300 includes simulating emission of RF signals, in the mmWave range, from the target entity within the digital attack environment.

[0079] At step 308, the process 300 includes calculating RF signals reflected back from objects in the digital attack environment. Generally, the reflected RF signals include signals reflected back from the passive mmWave attack tag. Put another way, the passive mmWave attack tag may be positioned within the digital environment and the effect of the passive mmWave attack tag on reflect RF signals is calculated or determined. In some implementations, the reflected RF signals are calculated using ray tracing techniques.

[0080] At step 310, the process 300 includes determining simulated time-domain RF signals from the reflected RF signals.

[0081] At step 312, the process 300 includes using the simulated time-domain RF signals to optimize parameters of the passive mmWave attack tag (e.g., using process 200, above).

[0082] Millimeter-wave (mmWave) sensing has been employed extensively in many critical applications (e.g., automotive vehicle

[0064] ,

[0062] , security

[0020] , and robot

[0092] ) due to its exceptional ability to determine the targets’ range, angle, and velocity, and works in harsh weather and climate conditions (e.g., lighting, smoke, and fog

[0085] ,

[0077] ). These radars have been used for obstacle detection, speed measurement, localization, invasion detection, blind-spot detection, collision avoidance, and surveillance. Due to their widespread use, particular emphasis was laid on investigating the various attacks on mmWave sensing (hereafter mmWave attack) and identifying solutions to mitigate these attacks. These attacks on mmWave sensing can interfere with the sensing measurements to disappear the ahead object or create the ghost object, which can directly or indirectly cause serious consequences such as crash accidents, maliciousinvasion, and sudden braking. Thus, these radar sensing units are prone to wireless sensing errors that can lead to severe consequences. For example, there is the infamous ghosting effect from perimeter security, robots, and automotive vehicles where it either detects something in front that does not exist (e.g., false alarm and phantom braking)

[0019] , or it misses something that should have been detected

[0018] ,

[0013] , causing undetected intrusion and accident collision. Such error- prone sensing mechanisms can further be designed for meticulously targeted attacks (i.e., attack on mmWave sensing).

[0083] Given the unique properties of the mmWave signal (loosely recognized from 24GHz- 300GHz), current attack vectors to mmWave devices rely on active electronic-based approaches (e.g., active jamming of electromagnetic waves and injecting malicious signals). However, such attack signals are generated from high-cost, and proprietary equipment, lending these attack vectors are highly expensive, easily detectable track provenience, and even preventable

[0080] ,

[0060] , On the other side, researchers have explored that non-electronic meta- material tags can passively modulate RF signals for communication and networking applications

[0032] ,

[0033] ,

[0031] , Metamaterial is a new class of functional materials designed around unique patterns or structures, which cause them to interact with RF signals and other forms of energy in ways not found in nature

[0093] ,

[0091] , Given these tag materials’ low cost and ubiquity, one can easily customize a meta-material tag that tampers the modulation of mmWave signals to launch more practical and pervasive attacks on a range of mmWave sensing applications. Motivated by the above, the present disclosure explores whether it is possible to use low-cost meta-material tags to incur high-consequence attacks on mmWave sensing.

[0084] In this disclosure, the design and implementation of a meta-material-enhanced tag attack on mmWave sensing (also referred to herein as “MetaWave”) is presented, for example, as described in connection with FIGS. 4A-4E.

[0085] FIG. 4A is a schematic diagram showing an example MetaWave framework. An mmWave sensing attack can be implemented with meta-material -enhanced tags (as shown, MetaWave tags 402). The tags 402 can be attached to objects as paper tags and then hide an obstacle or intruder and mislead the mmWave sensing measurements / readings 406 obtained via an mmWave sensor 404 to produce lethal consequences. The dashed line 406A in the sensor readings 406 shown in FIG. 4A is natural, and the solid line 406B represents tampered.

[0086] FIG. 4F is a schematic diagram showing an example MetaWave system. The MetaWave system comprises two main modules as shown in FIG. 4F: (i) MetaWave Tag uses the characteristics of the meta-material to modulate the mmWave signal passively. Specially designed tag patterns, meta-materials, and deployment methods make it possible to manipulateecho mmWave signal and attack radar measurement results toward the attack goal; (ii) MetaWave Attack Framework further advances the MetaWave tag via an RF simulator to optimize the attack. The MetaWave tag optimizer module iteratively fine-tunes and optimizes the tag design and deployment parameters in a simulated attack scenario to achieve optimal attack performance. In particular, MetaWave features include: (1) Stealth: unlike existing radio frequency (RF) attacks that reply using RF modulators to perform active attacks, MetaWave produces fully passive meta-material tags for stealthy on-site deployment, which exposes little information regarding the attacker; (2) Viability: MetaWave embodiment is to use low-cost and easily obtainable commercial off-the-shelf (COTS) materials with customized tag designs to attack mmWave sensing systems, which has a considerably low bar to launch attacks (simple as paper tags); (3) Versatility: MetaWave offers multi -function attacks through a united design framework (as shown in FIG. 4B-4E). Specifically, in some examples, MetaWave utilizes three common COTS materials: C-RAM LF, tin foil, and copper parallel wire grid to modulate and affect the victim’s mmWave sensing system passively. C-RAM LF is a type of flexible, broadband, lightweight radar absorbing material comprising graded layers of lossy open-cell plastic foam which can provide approximately 20 dB of reflectivity reduction in a particular frequency range. In addition, the MetaWave tag is optimized via a simulator to achieve optimal robustness and practicality.

[0087] To realize MetaWave, two technical challenges were addressed: (a) Design and implement easy-obtainable metamaterial tags for the mmWave sensing attack. The foundation of the MetaWave attack rests on the tag modulating signals from victims’ mmWave sensing systems. When the incidental mmWave signal meets the tag, the designated modulation effect will be triggered, and the victim will now receive the modified echo signal. Inspired by such modulation mechanisms, three different materials were recruited to attain three typical mmWave modulations (e.g., absorption, reflection, and polarization). With these tags, the attackers can realize ghost attack (GA) and vanish attack (VA), even on multiple objects simultaneously against the RF fingerprint protection, (b) Simulation model and optimization of the tag attack for robustness and practicality. Since the diffraction phenomenon of RF is more complex than that of light, the RF simulation is more demanding than the simulation of other sensing modalities (e.g., camera and lidar). Mainstream RF simulators are not practical in this disclosure due to their high computational overhead. Moreover, the material types, patterns, and deployment parameters (e.g., height, orientation) of the MetaWave tags jointly lead to a high-dimensional parameter space of the design. Finding the optimal parameters in this high-dimensional space that satisfies the demanding attack requirement is also a non-trivial challenge. To make the problem tractable,a specially designed Simulator was utilized to create a digital replica of the physical world using the inversible simulations of mmWave signals. Compared to mainstream simulators, the proposed simulator is capable of calculating the mmWave echo signal of arbitrary objects in an accurate and efficient manner. Then, a gradient descent approach is leveraged to find the optimal attack parameter solution iteratively.

[0088] That is, the system automatically outputs the MetaWave tag’s optimal material type, pattern, size, and location. Extensive real-world experiments show that MetaWave can achieve an average attack success rate of 97% on range estimation, 96% on angle estimation, and 91% on speed estimation.

[0089] Our contributions can be summarized in the following four points: (i) A new passive attack type with meta-material enhanced tags on mmWave sensing and investigation of the feasibility of these security threats, (ii) Design and development of the first low-cost and easily obtainable meta-material -enhanced tags with specific designs for mmWave ghost and vanish attacks, (iii) Design and implementation of a simulator-based mmWave attack framework to optimize the attack. It can enable stealthy and viable attacks that rapidly analyze the physical environment and generate MetaWave tag design for a trap setup. This work will be open- sourced. (iv) The system performance and robustness in representative mmWave sensing scenarios were extensively evaluated in both simulation and real-world. Further, the system is examined in actual practice scenarios and present countermeasures against these meta-material- enhanced tag attacks.

[0090] Threat Model

[0091] Consider an adversary Eve (Attacker) that attempts to spoof Bob (Victim), a victim that leverages mmWave sensing for measurement and detection, such as adaptive vehicle cruising and motion surveillance monitoring. Two targeted attack cases are considered: (1) Vanish Attack (VA), which either hides an obstacle that Bob should avoid or hide Eve from being detected while trespassing, and (2) Ghost Attack (GA), which makes objects appear out of thin air, which triggers false alerts for obstacle detection or trespassing security alerts. Attack examples are shown in FIG. 4B-4E which are a series of schematic diagrams each depicting a typical MetaWave attack scenario.

[0092] FIG. 4B-4E are schematic diagrams illustrating four different MetaWave attack scenarios. FIG. 4B shows a vanish attack (VA) example where Eve places a MetaWave tag to hide a hard obstacle that Bob should avoid for personal safety. FIG. 4C shows a VA example where Eve uses a MetaWave tag 402 to evade detection. FIG. 4D shows a ghost attack (GA)example where Eve uses MetaWave tag to trigger Bob’s vehicle’s auto-braking safety feature which may lead to rear-end collision from cars behind. FIG. 4E shows a GA example where Eve utilizes an intermediate RF relay to perform the multi-path attack.

[0093] In contrast to prior work, the following desirable features are envisioned when Eve practically launches an attack.

[0094] Passive Tag'. The active electromagnetic attack signal or electronic device can be detected by forensic tools or prevented by the security check (e.g., RF fingerprint), and the electromagnetic wave generation equipment is expensive, requiring professional operating knowledge. Therefore, it is hard for Eve to launch an attack using an active electromagnetic wave generator to interfere or jam with Eve’s wireless sensing system.

[0095] Practical'. Assuming Bob puts the mmWave radar and computing devices in a well- controlled environment (e.g., a locked garage or security check routinely). Therefore, it is difficult for Eve to access and modify the devices imperceptibly in advance.

[0096] Black Box'. Assuming Bob’s mmWave sensing system and the computing platform is isolated / secured from the Internet or any other communication channels. Although Eve can have some knowledge about mmWave sensing signal information (e.g., signal frequency and bandwidth), Eve does not have access to the param eters / details of the sensing algorithms implemented, thus, attempting black-box attacks.

[0097] To the inventors’ knowledge, none of the existing mmWave sensing attacks can work under an application scenario with the constraints mentioned above. In addition, tags can be flexibly attached to carrier devices for deployment in various attack scenarios: (1) Sticker: The tag can be disguised as a sticker by attaching it to the surface of the object. (2) Sign: The tag can be made into general common artifacts (e.g., traffic or roadside signs) to be placed near the target. (3) Drone: The tag can be hung under the drone to get close to the target. The weight of the drone can be less than 250g, and then the drone will not be regulated by the remote ID from Federal Aviation Administration

[0036] , The operating distance (i.e., the approximate distance from the attacker hidden to the victim) could be more than 200m. Therefore, the challenge at Bob’s hand is how to attack mmWave sensing with designed MetaWave tags without violating the constraints above. Considering the current mmWave sensors follow standards regulated by spectrum band licenses (e.g., 24GHz) and industrial mmWave sensor parameters are publicly accessible via administrative documents, patents, and media reports (e.g., Bosch

[0014] , Continental [9]), it is practical for attackers to get the sensing signal information of the victim mmWave sensor from open public sources.

[0098] Rationale and Preliminaries

[0099] mmWave Sensing Measurements'. There are multiple signal modulation technologies in mmWave sensing, such as Frequency-Modulated Continuous-Wave (FMCW), Continuous Wave (CW), and Frequency Modulated Shift Keying (FSK / FMSK). Considering FMCW is the most representative one in real practice and applied in prevailing sensing applications, FMCW is used as an illustrative example. A typical mmWave probe in FMCW mode has one transmitter and multiple receivers. The transmitter generates the chirp waveform as s(t) = cos (2nfstartt + nBt2), where fstart is the transmission start frequency, B is the sweep bandwidth (fend—T chirp fstart), and Tchirpis the chirp duration. Then the chirp signal is reflected back by the object at the distance d and received by the receivers at the probe. The received signal is given by r(t) = the time delay dueto an object at distance d moving with velocity u with respect to the probe, c is the speed of mmWave signal in air, and a represents the amplitude of the received signal. Thus, the received signal consists of three fundamental properties of the sensing target (i.e., range d, angle 9, and speed u) for measurement, which are the basis of the mmWave sensing.

[0100] Then the range estimation is derived by d = where Tcis the up-chirp time, andfbis the beat frequency corresponding to the target. The angle estimation is 0 = arcsin(— ), where A9 is the phase difference at the receive antennas, Z is the distance of the receive antennas, and X is the wavelength of the sensing single. The speed estimation is v = - , where A9 isthe phase shift.

[0101] Meta-material Tags: FIG. 4G is a schematic diagram 400G depicting targeted absorption, reflection, and polarization of incident mmWave signals which are essential characteristics of MetaWave tag design. The proposed MetaWave tags include negative refractive index (i.e., mmWave absorbing) meta-material, and RF modulations metamaterial (i.e., specular reflection and polarization) as shown in FIG. 4G. Rather than directly attacking the computing parts of a mmWave sensing system, MetaWave tag aims at directly modulating the sensing echo signal while physically present in the environment. Proper attack materials were selected that can achieve low-cost and high-precision attacks while being harmless to the human body and environment. It is worth noting that most of the meta-materials can cover a wide range of RF frequencies, enabling high robustness and scalability in real practices. For example, therefection tag (e.g., metal foil-based) can create a strong reflection for RF signal from a wide frequency band supporting Sub-6GHz and Terahertz bands

[0012] ,

[0065] ,

[0079] ,

[0102] Absorption Tag 402G is to largely attenuate the amplitude of the echo signal, which can make the probe believe there is no object detected and make the obstacle / intruder ahead disappear. The C-RAM LF material

[0016] was found to be the particularly suitable for attacking mmWave. C-RAM LF is layers of lossy open-cell plastic foam. It is lightweight, flexible, and can be bonded to many metals, plastic, or wooden surfaces using a polychloroprene contact adhesive. The frequency range of use is 18-40 GHz with -20dB. It has high service temperature of a maximum of 120° C. Its thermal conductivity is 6.5 X 10-5cal-cm / sec-cm2-° C and density is 0.07g / cm3. C-RAM LF is considered as a meta-material because it was not the materials (e.g., metal wires, silicon chips, etc.) traditionally utilized in RF circuit and system designs.

[0103] Reflection Tag 404G is to rebound the transmission signal and spoof the object detection by overwhelming the object signal and utilizing the range resolution limitation (round 0.75-0.9m in 24GHz probe)

[0060] , It can create ghost objects to the radar and affect the position and speed measurements. The tin foil is the most suitable for the reflection tag since it is malleable and can create a strong electromagnetic reflection. The high-frequency RF energy is strongly reflected when the frequency is higher than 100 MHz (e.g., 24-300GHz). The density of tin foil is 2.7g / cm2, but it only has a thickness of 0.016 mm, making it ultra-low-weight in practice.

[0104] Polarization Tag 406G is to restrict the specific fields in the mmWave signal movement since the mmWave signal is the transverse wave, which can tamper with the echo signal. The polarization tag is mainly used for performing GA for speed and angle measurements.Specifically designed parallel grid made of fine copper wires were used as polarization tags

[0015] , The polarization tag can attenuate EM waves in the parallel wire direction by more than 90% while having almost no effect on EM waves in the perpendicular wire direction. It is a combination of cardboard and copper wire, with a thickness of 3mm and a density of around 8g / cm3.

[0105] Simulator to Optimize Attack Designs: Different from the adversarial examples aiming for deep neural networks

[0078] , the physical attack prefers manipulating the physical environment to spoof and mislead the sensing measurements leading to disaster aftermath (e.g., a crash)

[0026] , Compared to other active attacks, this physical attack works passively and is harder to discover and trace with existing forensic tools (e.g., RF fingerprint and false alarm detection). In this disclosure, the designed simulator can provide a digital replica of the corresponding attack scenario (i.e., the physical object or process) and simulate the tag modulation effects on themmWave signal. The optimization concept was also utilized in the simulator

[0030] to optimize the tag design and deployment parameters (i.e., the physical manipulation) to achieve a practical and robust mmWave attack.

[0106] Experimental Results

[0107] To examine the feasibility of MetaWave attacks using the proposed meta-material tags, various studies were conducted including real-world VA feasibility studies using mmWave sensing probe and a comer reflector in the real world. In some studies, a trihedral comer reflector (30 cm height and width) was first placed four meters in front of the mmWave sensing probe, and its original mmWave response was recorded. FIG. 4H and FIG. 41 are graphs depicting ranged amplitude analysis using Fast Fourier Transform. FIG. 4H shows radar RX in line of sight with the comer reflector and FIG. 41 shows absorption tag placed in front of the corner reflector. As shown in FIG. 4H, a peak that represents the corner reflector is detected at four meters range using Fast Fourier transform (FFT). Then, a real mmWave absorbing metamaterial- enhanced tag was applied on the comer reflector and the attacked mmWave response was recorded. As shown in FIG. 41, the metamaterial absorbing tag is effectively erasing the response wave from the comer reflector. Making radar receivers disabled to sense the object fundamentally. Currently, such fundamental attacks are undetectable using mmWave sensing systems alone.

[0108] System Overview

[0109] MetaWave Tag Design-. As noted above, MetaWave’s meta-material tags are designed to be malleable and / or to adhere to various types of surfaces (e.g., moving or stationary objects). FIG. 4J shows example two-dimensional (2D) and three-dimensional (3D) MetaWave tags 410(a)-(m) designed to blend into various environments. Various representative tag patterns are illustrated. Tags 410(a)-(e) are basic 2D patterns, tags 410(f)-(h) are advanced 2D patterns, and tags 410(i)-(j) are 3D patterns. Tags 410(k)-(m) are real-world prototypes of absorption, reflection, and polarization tags. As shown in FIG. 4J, tags can be made into various shapes and structures to blend with an environment to improve stealthiness. Moreover, these meta-materials are COTS, making the tags easy to make.

[0110] After extensive evaluation, it was found that the hexagon is an optimal and widely applicable candidate for the basic tag element. Since the hexagon is densely paved and has reflection symmetry, the hexagon shape has less overlap than other shapes (e.g., circle) and can be composed into honeycomb patterns, such as the example tag 410(h) shown in FIG. 4 J. Inaddition, each honeycomb hexagon’s material and orientation can be individually adjusted and reused. Therefore, the honeycomb pattern provides overall robustness at an ultra-low-cost.[OHl] As shown in FIG. 4J, the MetaWave supports 3D patterns such as the example tags 410(i) and 410(j ). The 3D patterns have the unique advantage of multi-directional attack, capable of attacking in multiple directions with no impact of orientation. Moreover, the 3D patterns can be rotated to generate micro-motion for attacking the Doppler effect-based measurement algorithms, thus further improving the attack performance.Table I: Scene parameter examples in the proposed simulator

[0112] FIG. 4K is a schematic diagram depicting an example MetaWave attack framework. As shown the example MetaWave attack framework includes a differentiable RF simulator 402K and a MetaWave Tag Optimizer 404K. The system is configured to obtain scene parameters 406K (e.g., victim and trap parameters) as input in addition to, in some implementations, MetaWave tag parameters, and output an optimized tag design 41 OK with specific tag deployment parameters such as location. As shown, in various examples, the system is configured to automatically iteratively adjust tag parameters until, for example, a decision threshold is satisfied.

[0113] MetaWave Attack Framework Design: 1) Scene Parameters: As shown in FIG. 4K and Table I, two types of parameters are needed to define an attack scenario. The first one is thevictim parameters, including the victim’s radar sensor frequency, bandwidth, polarization mode, and sensing distance. These parameters are usually publicly open or can be obtained from the target product’s manual or even through social engineering, alternatively, by using an RF signal receiver to sample the target radar’s parameters. It is worth mentioning that existing RF modulator-based attack approaches also need these parameters to launch attacks (as detailed below and as shown in Table III). The second type is the trap parameters that include the geometry and material of the objects (e.g., walls, vehicles) and the trap environment. These objects usually have a consistent geometry, or their crucial descriptors are easily obtainable due to the manufacture standards and marketing or user needs, so common pre-defined geometries can be used. Also, the characterization of the scenario environment can be observable via remote measurements (e.g., telescope, augmented reality) or extracted through public navigation maps (e.g., Google Maps). The total number of scene parameters is about 1000, most of which define the object geometry. 2) Differentiable RF Simulator: To find the optimal design of MetaWave tags for a given scenario, a specially designed RF simulator is needed that can both output the simulated sensing signal and provide guidance for optimizing the tag design in the gradient estimation.

[0114] To achieve this, a simulation technique is needed that is (1) Differentiable: The differentiability of the simulator can dramatically increase the optimization speed of the gradientbased optimizer; (2) Efficient: the highly efficient simulator can enable fast iteration for finding optimal tag design. However, existing Computational Electromagnetics methods such as Finite Element Method (FEM)

[0046] and Method of Moments (MoM)

[0038] have huge computational costs and are therefore unsuitable for fast iteration and hence are not suitable for this work. Ray tracing is a method of computer graphics that simulates the physical behavior of light. Since RF and light are both electromagnetic radiations, the ray-tracing method can also be applied to RF. MetaWave’s simulation is based on the Shooting and Bouncing Rays (SBR)

[0056] , The ray tracing-based method is much faster than other RF simulation methods (e.g., FEM

[0046] and MoM

[0038] ).

[0115] RF -Object Intersection : To simulate the interaction of RF signals with the object geometry, it is necessary to find the specific location where RF intersects the object. And because of the wave-particle duality of EM and to reduce the computational complexity, only the particle properties of the RF are considered at this stage and emitted as a ray. The geometry of the object in the digital world is usually represented by thousands to millions of triangles (i.e., mesh). Since direct intersection computation of mesh is challenging, the intersection detection is split with mesh into intersection detection with each triangle within its mesh and calculate theintegral of all intersected rays to get approximate solutions. The whole process is called ray tracing.

[0116] The Bounding Volume Hierarchy (BVH)

[0082] structure was utilized to store the triangles, which is an optimization structure-based binary search tree that has the logarithmic time complexity of finding the nearest triangle at a given location.

[0117] The Meoller-Trumbore intersection algorithm

[0059] was used for the fast ray-triangle intersection calculation. The Monte Carlo method is used to calculate the approximation of all rays emitted and received by the radar. However, the classic random sampler in ray tracing and SBR is considered indifferentiable. Therefore, a new edge sampler

[0051] is applied in MetaWave’s sampling state, which is continuous and can be differentiated.

[0118] Material Simulation: Material simulation and definition are essential for the attack system because of the need to simulate MetaWave tags with different RF modulation methods. However, the native SBR method does not support the simulation of materials. Therefore, SBR is enhanced by introducing an extension of the Bidirectional Scattering Distribution Function (BSDF) in the ray-tracing stage.

[0119] BSDF is a universal function that defines an arbitrary surface’s reflection and refraction properties

[0024] , as shown in FIG. 4G. The BSDF takes incident ray and reflection ray direction as input and returns the probability of the reflection happening. In addition, it can also change the characteristics of the rays (e.g., polarization).

[0120] RCS Estimation: The Radar Cross-section (RCS) of simulated objects is estimated via Physical Optics Integral (POI). The POI is used to simulate the interference, diffraction and polarization of EM wave propagation

[0021] ,

[0056] ,

[0022] ,

[0037] , The scattered far-field is calculated by:

[0121] Where (p and 0 are the unit elevation vector and unit azimuth vector. k0is the wave number. Aeand A(pcan be calculated by:

[0122] Where E(ri') and H(ri) are the electric field and the magnetic field, n is the surface normal, k is the unit vector of ray direction. Zois the intrinsic impedance. A.4 is the area of ray tube projection. rtis the ithreflected point.

[0123] FMCW Propagation'. It is worth mentioning that the developed MetaWave simulator can work for various signal modulation technologies (e.g., FMCW, CW) in signal propagation. As discussed herein, FMCW is used as an illustrative example. The FMCW Propagation module first generates an FMCW waveform from digital radar. Then, the transmitted signal is manipulated according to the response of the objects. The echo signal is denoted as:

[0124] where x is the incoming signal, y is the response echo signal, and X is the wavelength of the incoming signal. It is worth noting that X can be configured to simulate RF with different frequencies.

[0125] The simulator takes the scene parameters as input. A digital copy of the attack trap is created based on the parameters of environmental, objects, and radar specifications. The digital radar then emits an RF signal, which is received by the radar and approximated by a number of ray samples. Each ray is passed through the RF-Object Intersection module to calculate the ray exposure point, and then the reflected / refracted rays are calculated based on the surface properties of the Material Simulation module. The RCS values are then fed into the FMCW Propagation module to calculate the echo signal for FMCW. The simulator outputs the timedomain RF signal.

[0126] To illustrate the fidelity and effectiveness of this simulator, the simulated signal was compared with other commercial simulator products described herein. To reduce the effect of noise in the time domain signals, the signals were converted to the frequency domain via an FFT heatmap. Structural similarity (SSIM)

[0083] was used to measure the similarity. SSIM is a full reference method (FR)

[0076] that considers both structural information and perceptual phenomena of the data. It is better at handling two-dimensional signal data than metrics such as peak signal- to-noise ratio (PSNR)

[0075] , The SSIM outputs a value range from 0 to 1. The higher the SSIM, the better the similarity.

[0127] MetaWave Tag Optimizer'. Given a scenario 5, the attacker aims to design a MetaWave tag such that when deployed in this scenario, the mmWave measurement results are manipulated according to the attack goal. The MetaWave attack on mmWave sensing via MetaWave tag can be formulated as:

[0128] where x is the tag parameters, y is the attack goal value, f is a mmWave measurement algorithm. It is worth noting that MetaWave only requires querying these measurement algorithms and does not need implementation details. S is the RF simulator, S(<5 + x) is the simulated mmWave signal of the scenario with MetaWave tag, <J is the attack success criterion threshold. The attack goal is realized when the loss function L < <J is achieved. The attack goal value y is the mmWave measurement results of the ideal successful attack scenario, y for disappearing attack is set to 0 since the attack excepted the target disappeared on the radar, y for the creation attack is set to the detection threshold of the target measurement algorithm. To minimize the L2 norm in Equation 4, the natural evolution strategy (NES)

[0042] was employed, automatic differentiation (AD) was used for gradient estimation, and Stochastic Gradient Descent (SGD)

[0048] ,

[0074] was used as an optimizer to fine-tune the MetaWave tags.

[0129] Attack System Integration

[0130] Meta Wave Tag Advancement: In this section, the MetaWave tag design and deployment parameters are further advanced using the simulator. The simulator can automatically optimize the MetaWave tag design for different attack scenarios. To simulate the meta-materials, BSDF is defined for each material according to its characteristics.Absorption Tag in MetaWave Simulator '. The C-RAM LF material has a -20db attenuation for mmWave signals, meaning that it is able to reduce energy reflections by 90%. Therefore, for the BSDF of absorbing material, a 90% probability of absorbing rays and a 10% probability of Lambertian reflection is set. The BSDF of C-RAM LF material is denoted as:, where (0 , 0f) is direction of incident rays, (0r, 0r) is direction of reflection rays. psis the albedo of the material. In this disclosure, ps= 0.1.

[0131] Re flection Tag in Meta Wave Simulator '. Tin foil tape was used as the reflection material. The reflection of metal for millimeter wave follows the same specular reflection as visible light. The probability of ray reflection is related to the angle of incidence and the angle of the normal plus fuzziness. The BSDF of specular reflection is denoted as BSDR(0t, 0i, 6r, 0r) = ps6 0i — 0^6^01 + 7i — 0V), where (0,, 0t) is direction of incident rays, (0r, 0r) is direction of reflection rays. psis the albedo of the material, in this disclosure, psis the specular albedo of the material, 5 is the dirac delta function.

[0132] Polarization Tag in MetaWave Simulator '. mmWaves waves pass directly through linearly polarizing materials and are not reflected or scattered. However, the material changes the way mmWave is polarized. The material retains the polarization in the tag’s tangent directionand rejects the polarization in the bitangent direction. As shown in FIG. 4G (polarization tag 406G), a circularly polarized mmWave will transform to linearly polarized after passing through.

[0133] Simulator -based MetaWave attack on mmWave sensing: Different from most adversarial attacks that are focusing on the computing module (e.g., the target recognition / classification algorithms), MetaWave targets the physical characteristic of RF via a differentiable RF simulator.

[0134] FIG. 4L is a schematic diagram depicting an example MetaWave simulator 400L in which scene parameters 406L are defined in three parts to characterize an attack scenario: tag parameters, victim parameters, and trap parameters. For instance, the MetaWave tag vanishing the roadblock can lead to a severe collision on highways with roadworks. The MetaWave Tag optimizer 404L tunes tag parameters until the victim’s mmWave sensor 403L makes a mistake. As noted above, MetaWave’s simulator-based attack system utilizes an iterative approach to find the optimal tag design 410L and deployment parameters. The integrated end-to-end attack workflow is illustrated in Algorithm 1 below and can be described as follow:

[0135] The attacker first defines an attack scenario using scene parameters. MetaWave creates a MetaWave tag in the simulator with randomly initialized tag parameters as shown in Table II below. Then, the MetaWave framework iteratively fine-tunes and optimizes the tag parameters to achieve the best attack performance and robustness. When the attacking loss is less than a predefined threshold or the maximum iterations count is reached, MetaWave outputs the tag design and deployment parameters. The attacker creates an actual MetaWave tag and deploys it following these parameters. The tag parameters are designed to have fault tolerance for real- world deployments to improve feasibility.

[0136] In this disclosure, three meta-materials are defined as mentioned above. These metamaterial tags can also be optimized to work for different frequencies simultaneously. Besides, it is worth noting that MetaWave is a general mmWave attack framework since the target algorithm, tag materials, etc., are configurable. In other words, it is highly scalable and supports wide range of RF frequencies and various sensing measurement algorithms and meta-materials.Table II: Tag parameter examples for mmWave attack

[0137] Algorithm 1 MetaWave Attack Framework

[0138] Input: sceneP aram: Attack scenario parameters

[0139] radar: Radar setting

[0140] RFmeasure: RF measurement Algorithm

[0141] goal: Attack Goal Value

[0142] Output: tagParam: MetaWave tag parameter.

[0143] 1 : tarParam = TagParam.randomInit()

[0144] 2 : sceneGen = SceneGenerator.create(sceneParam)

[0145] 3: simulator = DiffRFSim.create()

[0146] 4: optimizer = Optimizer. create()

[0147] 5 : for in range(MAX ITER) do:

[0148] 6 : scene = sceneGen. generate(scene, tagParam)

[0149] 7 : simSignal = simulator. sim(scene)

[0150] 8 : simValue = RFmeasure(simSignal)

[0151] 9 : loss = Loss(simValue, goal)

[0152] 10: if loss < threshold then

[0153] 11 : return tagParam

[0154] 12: end if

[0155] 13: optimizer. zeroGrad()

[0156] 14: loss.backward()

[0157] 15: optimizer. step()

[0158] 16: end for

[0159] 17: return tagParam

[0160] Evaluation Setup. MetaWave’s simulation and real -world attack performance were evaluated.

[0161] mmWave Probe'. A representative and commercial off-the-shelf radar was used with 24Ghz carrier frequency and FMCW modulation

[0044] , The supply voltage varies from 3.3V to 5.5V, the size of the probe is 5.0cm x 4.5cm and the weight of the probe is 250g, and the cost of the probe is less than $300. Thus, this probe is suitable as a testbed to be spoofed.

[0162] Computing Device : MetaWave’s system is completed on an ordinary PC equipped with an Intel Core i7-8700K 3.70GHz CPU and one NVIDIA GTX1080 8 GB GPU.

[0163] Environments'. MetaWave was evaluated in 20 different environments including indoor and outdoor parking lots, roads, and traffic intersections, at different times of days (8:00 - 22:00) with different weather conditions (e.g., fog, snow, and wind). FIG. 6 is a series of images (a)-(h) showing various environments at different times that were utilized in some of the studies.Specifically, (a,b,e,d,g) are Roadsides, (c, f, h) are parking lots, (a,b,d,h) are during daytime, (c,e,f,g) are during nighttime, and (d,h) are during snowy weather.

[0164] Objects Preparation'. 17 different objects that most often appear in sensing scenarios were recruited, including roadblocks, humans, rocks, garbage cans, trees, and cars. All procedures follow the Institutional Review Board (IRB) protocol.

[0165] FIG. 5A shows an example system implementation including the designed metamaterial tags (from left to right: absorption tag 502A, reflection tag 502B, and polarization tag 502C) with a mmWave sensing probe 504 mounted on a vehicle / stand and the obstacle 506, imitating the mmWave sensor working in the real world.

[0166] Data Acquisition'. As shown in FIG. 5 A, COTS mmWave radar’s sensing signal was collected from the victim’s perspective. For each trial, a 60s raw signal was collected from victim’s radar to evaluate if the success rate that sensing result is spoofed. The obstacle 506 is deployed at a four-meter range from the mmWave radar by default. The tags’ deployment (tags 502A, 502B, 502C) varies by application and will be specified for each experiment (e.g., on the obstacle, non-contact, hung on a drone, etc.). The tags 502A, 502B, 502C are hexagon-shaped,and the size is according to the attack system’s optimal output. The tag size ranges from 6cm to 50cm in edge length. In addition, square shaped tags with 100cm edge length were tested.

[0167] Evaluation Metrics: To measure the effectiveness, pervasiveness, and adaptability of MetaWave attack in the real world, Top-k is employed for attack success rate. Top-k is a probabilistic data structure that can be used to identify frequently and / or infrequently occurring instances (e.g., items) in a data stream or corpus. In this study, k = 1, 3, and 5, which means we attack 1, 3, and 5 times for the same attack scenario. If there is any successful attack, the whole trial attack is considered successful. Top-k inference accuracy is defined as the percentage of successful trials. As discussed herein, the range, angle, and speed are the three fundamental properties of the sensing target. To prove MetaWave attack is not due to victim sensors’ hardware measurement error-tolerance, a stricter and more reasonable attack success criterion is defined that should be larger than the tolerance range

[0043] : (1) Range Measurement: the attack is considered a success if the sensing algorithm output no detection while a real object is presented, or vice versa (or more than ±0.9m from the actual range). (2) Angle Measurement: attack success if the sensing output is more than ±10° from the actual angle. (3) Speed Measurement: attack success if the sensing output is larger than ±0.19m / s from the actual speed. The attack result is quantified by calculating the Average Misalignment (AM) of the victim’s measurement. Each Misalignment is calculated by — — * 100%, where x is the sensing measurement result, y is the ground truth, and <J is the normal measurement tolerance. If AM is larger than 100%, it implies a successful attack. Standard deviation (STD) is also adopted on top of AM to show case the attack robustness. With larger AM and STD, victim sensor experiences higher deviation from correct reading, leading to more severe consequences.

[0168] Simulator-Based Attack System Evaluation

[0169] System Simulation Performance: MetaWave simulator’s capability and effectiveness of RF propagation with meta-material tag were evaluated in various studies. The simulator’s result was compared with the real-world echo signal. FIG. 5B is a table showing results of a comparison between the real-world echo signal and the designed RF simulator with three representative mmWave sensing objects without (a)-(c) and with (e)-(g) tag attacks.

[0170] As shown in FIG. 5B, the MetaWave system was tested with the three most common objects in the attack scenarios (e.g., human, car, and roadblock) with and without attacks. The results are shown in range-FFT. The X-axis is the range, and the Y-axis is the amplitude of the spectral signal. The average similarity between the simulator result and the corresponding real signal is 75.2%. For all scenarios, the simulator described herein fully reflects the echo signalchange and outputs the same attack performance as the real-world one. Thereby, it was established that the simulator in MetaWave has an excellent ability to help us explore the metamaterials effect on RF signals.

[0171] Computational Overhead. The complexity of MetaWave simulator is approximately O(nlogri) where n is the number of triangles representing scene geometries. It is much faster than other simulation methods such as FEM

[0046] and MoM

[0038] , where complexity is O(n2). The working time of three representative objects with six configurations were evaluated in FIG. 5B. The average time cost for MetaWave to simulate mmWave echo signal for one scene is about Is, 80-100X faster than the representative and professional product, FeKO

[0017] ,

[0172] Attack Robustness in the Simulator: To improve the MetaWave tags’ deployment error tolerance and robustness, the location variation was included in the simulator-based optimization. Specifically, for each iteration, the MetaWave tag is placed at a distance of ± 15 cm and an angle of ± 10 degrees from the target location. The simulated VA attack results are shown in FIG. 5C which depicts results of robustness evaluation with the deployment condition in the simulator on tag orientation and tag position, respectively. It was observed that the simulated attack success rate remains high (above 99%) when radar orientation and tag orientation variation is less than 10 degrees. The average attack success rate in the simulator is 91.2%. The results are close to the performance in actual practice. Thereby, MetaWave can simulate the variation of the tag parameters in the simulator and then support multi-directional attacks.

[0173] Practicality and Generalization

[0174] Evaluation-. In this section, MetaWave’s attack performance for different attack goals and tasks in the real world were evaluated. Especially, for range, angle, and speed measurement attacks, the most representative and essential corresponding measurements in real practice are employed.

[0175] Overall Performance '. The ability of MetaWave to attack range, angle, and speed measurements with Top-1 / 3 / 5 inference accuracy were evaluated. FIG. 7A shows overall performance of attacking range, angle, and speed sensing measurements. As shown in FIG. 7A, the Top-1 attack success rate for three measurements are all above 90% (i.e., Range: 97%, Angle: 96%, Speed: 91%), achieving equivalent performance as other state-of-the-art mmWave attack approaches in a more low-cost way

[0080] , Besides, the attacks can cause far greater errors in measurement results than the typical measurement tolerance. When attacking the mmWave system’s range sensing, MetaWave can spoof the system and cause an average absolute error of1.3m (STD: 1.27m) on sensing results (AM 144%, STD 141%). When attacking angle measurements, MetaWave can cause an average absolute error of 38 degrees (STD: 14 degrees) on sensing results (AM 384%, STD 194%). For speed measurements, MetaWave can cause an maximum absolute error of 5.4 m / s (50 percentile: 2.5m / s) on sensing results (AM 1570%, STD 7368%).

[0176] MetaWave Tag Optimizer Analysis'. To reflect simulator optimization and improve MetaWave attack performance, the attacks were tested with and without the guidance in the simulator-based optimization. When deploying without the tag optimization, tags were set up based on manual experience and understanding of the attack mechanism (i.e., estimate the mmWave sensor positions and the mmWave signal propagation physically present in the environment). To avoid personal preference, the attacks were conducted without the tag optimization twice for each setting. FIG. 7B shows experimental results demonstrating the impact of the MetaWave Tag Optimizer on attack performance on range and angle measurements. The first column is the normal range reading without attack, the second and third columns are tag attacks without the optimizer, and the fourth column is the tag attacks with the optimizer. The tag attacks with the optimization achieve over 99% attack success rate, while others fail. As shown in FIG. 7B, both manual attacks without the optimization on both range and angle measurements lead to minute deviation from the normal one (the Top-1 attack success rates are all below 5%). In contrast, the attacks with the optimization lead to significant advancements (the Top-1 attack success rates are both over 99%). Overall, the tag optimizer enables attacks on the range and angle measurements and significantly improves their attack performance.

[0177] Trap Variation Analysis

[0178] Environmental Dynamic. To reflect the system performance withstanding environmental inferences in actual practice, four common environmental dynamic factors in daily life were considered for tests: (1) Temperature, range from -1°C to 20 °C; (2) Humidity controlled from 20% to 70% (3) Lighting, the light intensity is controlled from OLux to lOOOLux; (4) Magnetic field strength is controlled from 100T to 400T. FIG. 7C is a graph showing performance under different environmental dynamics, including temperature, humidity, lighting, and magnetism respectively. FIG. 7C demonstrates that the attack success rate is all above 95%. Thereby, MetaWave presents a solid tolerance to different environmental dynamics. FIG. 7D is a graph showing the longitude evaluation of a span of 12 months. FIG. 7E is a graph showing performance under different tag sizes.

[0179] Longitude -. To illustrate the reliability of the MetaWave tag, the longitude test was evaluated in 12 months. The MetaWave tags are stored in a normal environment with no special protective treatment. As shown in FIG. 7D, MetaWave can maintain a consistent and high attack success rate (> 95%) for an extended period.

[0180] Tag Size -. To illustrate the system performance under the impact of tag size, the system effectiveness is evaluated in the same application with difference tag size. In order to achieve the attacks, the tag size can be minimized to 20%-40% of the sensing cross-sectional area of the target objects. As shown in FIG. 7E, MetaWave attack size has better performance on larger tag, but still keep success rate even on smallest tags (>90% on 10cm length).

[0181] Attack Range Measurement: MetaWave’s attack success rate was evaluated based on tag distance. The range FFT

[0041] was used as the target distance measurement algorithm.

[0182] Impact of Tag Position'. In a real attack, the position of the tag relative to the radar may not exactly match the ideal deployment in the developed simulation. Therefore, MetaWave’s tolerance on the tag’s position was evaluated. FIG. 7F, FIG. 7G, and FIG. 7H depict results from evaluation of attacks on the range, angle, and speed measurements, respectively. The X-axis is the variance of tag position, the Y-axis is the variance of tag orientation, the color illustrates the attack success rate. As shown in FIG. 7F, the attack results reach a high (above 94%) success rate when the position variation is less than 15cm, and it remains above 90% when larger than 10cm. The results show MetaWave has strong tolerance on tag deploy position when attacking distance measurement.

[0183] Impact of Tag Orientation'. In a real attack, the orientation of the tag toward the radar may not exactly match the ideal deployment in the presented simulation. Therefore, MetaWave’s tolerance on the tag’s orientation is evaluated as shown in FIG. 7F. MetaWave gets a high (above 94%) attack success rate when the rotation variation is less than 7°, and it remains above 90% when the rotation variation is at 12°. FIG. 7F manifests that the impact of the orientation is higher than the range since the mmWave signal has narrow beamforming. The results show MetaWave has strong tolerance on tag deploy orientation when attacking range measurement.

[0184] Impact of Tag Attack Duration'. The stability of MetaWave’s attack is crucial for the feasibility of real attacks. Therefore, the system was tested with time duration from 0.3s to 1.5s. The average success rate is 97%, and all kept above 92%, which proves the reliability of MetaWave in the attack. The results show MetaWave is capable of attacking distance measurement with long-term stability.

[0185] Atack Angle Measurement: In this section, MetaWave’s performance when attacking angle measurement in real practice is evaluated. The Angle of Arrival (AoA)

[0047] was used as the target angle measurement algorithm. The results are shown in FIG. 7G.

[0186] Impact of Tag Position'. As shown in FIG. 7G, the attack results reach a high (above 95%) success rate when the position variation is less than 10cm, and it remains above 90% when larger than 10cm. Thereby, MetaWave has strong tolerance on tag deploy position when attacking angle measurement.

[0187] Impact of Tag Orientation'. MetaWave get a high (above 94%) attack success rate when the rotation variation is less than 7°, and it remains above 90% when the rotation variation is at 12°. It also shows the same phenomenon mentioned above due to the narrow beamforming of the mmWave signal. The results show when attacking angle measurement, MetaWave can keep high performance when tag deploy orientation is slightly different from ideal deployment.

[0188] Impact of Tag Attack Duration'. The system was tested with time duration from 0.3 s to 1.5s. The average success rate is 96%, and all kept above 88%, which proves MetaWave attack system is capable of attacking angle measurement with long term stability.

[0189] Attack Speed Measurement: mmWave sensing senses the velocity of the target object via the Doppler effect. In this disclosure, the Range-doppler algorithm

[0086] was used as the target algorithm. The results are shown in FIG. 7H. Unlike the distance and angle measurements, the attacks on speed measurement show irregular results because the doppler algorithm is susceptible to micro-motion.

[0190] Impact of Tag Position'. As shown in FIG. 7H, the attack results reach a high (above 92%) success rate when the position variation is less than 10cm, and it remains above 88% when larger than 10cm. The results show MetaWave has a strong tolerance on tag deploy position when attacking speed measurement in real practice.

[0191] Impact of Tag Orientation'. MetaWave get a high (above 94%) attack success rate when the rotation variation is less than 7°, and it remains above 90% when the rotation variation is at 12°. The minimum is 84% when the rotation error is larger than 20°. Thus, MetaWave shows a strong tolerance on tag deploy orientation when attacking speed measurement in real practice.

[0192] Impact of Tag Attack Duration'. The system was tested with time duration from 0.3 s to 1.5s. The average success rate is 91% and all kept above 85%. The average success rate is 97% while 5 frames have the best performance. At all time intervals, the average success rate kept above 73%, which proves MetaWave attack system can attack the speed measurement with longterm stability.

[0193] Real -World Attack Evaluation

[0194] The attack system can be utilized in various public spaces as a result of the high accessibility and portability of the setup. Thus, to evaluate MetaWave’s attack performance in the real world, the MetaWave attacks were conducted on mmWave sensing in five scenarios. In Scenario 1, 3, and 4, the mmWave sensor was mounted in front of a sedan car, and in Scenario 2, the mmWave sensor is deployed close to the security fence, imitating / hypothesizing the mmWave radar working on the vehicle / robot and the perimeter security, respectively, in the real world (but not on the real professional products).

[0195] FIG. 71 are experimental results showing attacked sensing results compared with normal sensing results without MetaWave spoofing. FIG. 8A shows an experimental setup for the multi-object attack. FIG. 8B and FIG. 8C are range FFT heatmaps showing the disappearance of subject 2’s trajectory (see FIG. 8 A) with the MetaWave tag. FIG. 8D and FIG. 8E are experimental results showing attack results when a victim is carrying an mmWave Sensor while moving. FIG. 8F shows operational examples depicting MetaWave tags disguised to improve stealthiness.

[0196] Scenario 1 (VA) : MetaWave tag’s ability to hide potentially dangerous objects (e.g., rock on the road) from mmWave sensing systems was tested. A rock was placed in front of the mmWave probe at the 3m range and use the MetaWave tag to cover it, as shown in FIG. 71(a). The average attack success rate is 88% (refer to FIG. 7J(a)), and the mmWave sensing system barely recognizes the obstacle continuously.

[0197] Scenario 2 (VA) : MetaWave can also be used to hide trespassing intruders from being detected by mmWave sensing systems. A mmWave surveillance probe was placed at 2 m height on a light pole to cover a rectangular-shaped perimeter as shown in FIG. 71(b). FIG. 7J(b) shows that MetaWave attack system successfully attacks the radar algorithm as it fails to detect the intruder with the attack success being 93 %.

[0198] Scenario 3 (VA) : MetaWave tag’s ability to spoof mmWave sensing systems in obstacle detection was tested. A MetaWave tag was placed on the roadside disguised as a normal house rental sign to make the pedestrian disappear in mmWave sensor sight, which can lead to vehicles hitting the pedestrian unexpectedly. The MetaWave tag is placed 0.5m away from the curb line, and a pedestrian walks in the vehicle’s direction right in front of the vehicle, as shown in FIG. 71(c). FIG. 7J(c) shows that the vehicle’s mmWave sensing system only recognizes the road sign around 40° on the side and filters out the pedestrian’s signal. The top-1 attack success rate is89%, which proves MetaWave can be extremely dangerous when an adversary deploys such attacks in real-life.

[0199] Scenario 4 (GA): MetaWave tag’s ability to create a ghost in front of mmWave sensing radar was examined. A 3D tag hung under a drone was deployed that creates a ghost obstacle before the mmWave probe. The attacker can constantly control and maintain the distance (around 1.5m here) between the vehicle and the tag via the drone

[0034] , as shown in FIG. 71(d). As a comparison, an actual obstacle moves at a speed of 0.5 meters per second. The tag can spoof the vehicle that there is only one static large object ahead and make it sudden brake or force it to change the lane. As shown in FIG. 7J(d), the average success rate is 95%, proving MetaWave has the capability to create the ghost object remotely.

[0200] Scenario 5 (VA): Multi-object Attack under the Complex Scene. MetaWave is also capable of targeting specific attacks in multi-object scenarios. As shown in FIG. 8B, the range FFT heatmap can clearly distinguish two subjects. However, the echo signal from Subject 2 is significantly decreased to ground noise level if Subject 2 is carrying a MetaWave tag, as shown in the lower part. It is worth mentioning that vanishing Subject 2 does not affect Subject l’s sensor reading, making the attack stealthy and difficult to recognize. Thereby, MetaWave is capable of the precise and practical multi -object attack.

[0201] Scenario 6 (VA & GA): Dynamic Attack of the Moving mmWave Sensor. MetaWave tag’s threat was examined in a dynamic scenario when the radar is moving toward the target at a speed of Im / s. As shown in FIG. 8D and FIG. 8E, MetaWave tag can alter the mmWave sensor’s reading in terms of both range (e.g., from 9m distance to disappear) and angle (e.g., from straight ahead to on the side), which spoofs the victim’s ability to detect forward collision. For attacking range measurement, the tag is attached to the target. For attacking angle measurement, the tag is placed in the middle of the vehicle and the object and at a 45 -degree angle from the vehicle’s initial position. As the vehicle moves forward, it passes the tag, so only the success rate of the attack is calculated before it passes by. The average success rate for range and angle (before pass by) is 92% and 93%, respectively. The results prove MetaWave has the capability to achieve GA and VA when the radar is moving.

[0202] Countermeasures. In this section, defense mechanisms against MetaWave from both security check (i - ii) and victim awareness perspectives (iii - v) are described.

[0203] (i) Manual Check: Civilians can take precautions by calling attention to suspicious people or signs around them, or even acquire law enforcement teams for inspection. However, such method is rarely effective given MetaWave tag’s ability to hide and camouflage as shownin FIG. 8F. Moreover, manual check is time-consuming (i.e., the forensics team can take minutes, hours to travel, while the mmWave sensing systems make decisions in real-time), and costly (i.e., giving up work duties and spending a long time on an object that only appears suspicious is not socially acceptable).

[0204] (ii) Radiation Detection: Another famous defense against RF attacks is to detect suspicious electromagnetic signals or radiation devices in the environment, similar to highway patrol officers detecting radar jammers

[0067] , However, the MetaWave tags work passively. The MetaWave tags were scanned with an Electromagnetic Fields (EMF) Radiation Detector product [4], and the scan result is illustrated in FIG. 8F, showing the MetaWave tags have nearly no active radiation and therefore struggle to prevent our attacks.

[0205] (iii) RF Fingerprint: Recently RF signal fingerprinting has gained attention in detecting sensing attacks, which utilizes the unique physical characteristics of the probe components to judge if the echo signal comes from the same electronic instruments. However, MetaWave directly attacks the sensing signal from the original mmWave probe instead of the computing part, making this countermeasure ineffective. This protection was evaluated with the approach in

[0080] , In the feature extraction, the statistical features of standard deviation, kurtosis, skewness of magnitude, and phase of the received signal are employed. Besides, the support vector machine is selected for one-class classification. As a result, the feature patterns between with / without attacks are similar, and the detection classification result is close to a random guess. Thus, this approach can not recognize the attack from MetaWave tags.

[0206] (iv) False Alarm Detection: In RF sensing practice, the Constant False-Alarm Rate (CFAR) technology

[0063] with the dynamic threshold is always used to rule out the active or passive noise and then resist the interference. However, meta-material tags were directly utilized to change the echo signal along with the environment as described herein. The classic CFAR detection [3],

[0073] was applied on the range-FFT of the echo signal with the setup in FIG. 5A, and it is also undetected in the VA attack and falsely detected in the GA attack. Thus, such false alarm detection could not resist our attacks.

[0207] (v) Multiply mmWave Sensors: It is possible that some devices may employ different mmWave sensors operating under different sensing frequency. However, MetaWave attacks are effective over a wide band of sensing frequencies simultaneously (e.g., multi -Ghz). Further, MetaWave tags can be inexpensively designated to specific ranges of frequencies and work together against redundant frequencies. Thus, this redundancy alone is unlikely to work.

[0208] Stealthy Analysis: The MetaWave tag works passively and can be camouflaged by plastic signs or billboards as common objects without causing visual vigilance (See FIG. 8F). Such camouflaging material (e.g., plastic, wood) often exhibits little interference to attack performance, making it deceitful for human visual inspection.

[0209] Sensing Modalities : The scope of this disclosure focuses on the mmWave sensing attack by modulating the physical echo signal under representative attack scenarios where the mmWave sensor has been widely used for years in practice. For example, some perimeter surveillance systems are based on mmWave sensing [6], [7], And some driver assistance systems’ adaptive cruise control functionality on automotive vehicles often solely rely on a frontfacing mmWave radar [8], [5], Besides, vehicles or robots can employ a variety of sensors (e.g., Lidar, Camera, mmWave) to detect the surrounding environment by making decisions together in real practice. However, due to the mmWave sensing system’s innate spatial perception advantages, applications’ high-level decisions rely heavily on mmWave for range, angle, and velocity measurements or under undesirable ambient conditions (e.g., lighting, smoke, and fog) where other sensors do not work well

[0069] ,

[0077] , Moreover, since the mmWave sensing result is fused in the final decision, the final decision is inevitably influenced if the mmWave sensing is impacted

[0085] ,

[0210] Attack Cost Analysis'. The average cost is $10-30 for the absorbing tag, $9-16 for the polarizing tag, and about two cents for the reflecting tag. Comparing to existing mmWave attack solutions as referenced in Table III, MetaWave is 10-100 times cheaper

[0061] ,Table III: A comparison of existing mmWave attack approaches. Note: these prior works do not provide the quantitative results

[0211] Practicality Analysis'. MetaWave can achieve a practical mmWave sensing attack with minimal cost. Overall, MetaWave can stealthily achieve a high attack success rate and withstand harsh weather. Besides, MetaWave’s tag can be integrated with mobile robots (e.g., drones, robotarms) for more complex attacks that involve motion for timely evasion and camouflage. Additionally, real-world environments were imitated with the mmWave radars in evaluations without justifying this attack on professional products, this work arouses public attention and provides in-depth explorations on this new attack type.

[0212] Attack Distance '. The mmWave radar in practical applications usually cannot precisely sense within around 0.2- Im because of the physical limitations. Therefore, for VA, the MetaWave tag can work when the distance is larger than 0.2m. The maximum sensing distance of the vehicle mmWave radar is usually 160m

[0010] , while the response distance to make the vehicle brake automatically depends on the speed of the vehicle (e.g., 73m stopping distance for a speed of 96km per hour

[0011] ). By optimizing the MetaWave tag, methods for vanishing obstacles or generating ghosts at the location of the maximum sensing distance are provided to achieve the attack.

[0213] Signal Modulation Technology. As mentioned herein, besides FMCW, MetaWave can also work on other popular types of signal modulation technology, such as CW, FSK / FMSK, and Digital Code Modulation (DCM). MetaWave can integrate different signal modulation technologies into the physical law-based calculation. Besides, the MetaWave attack mechanism is primarily based on the fundamental properties of the sensing signal.

[0214] Future Applications: (i) Since MetaWave modulates fundamental properties (e.g., amplitude, phase) of the sensing echo signal physically, MetaWave can also work on other developing and emerging mmWave applications or products (e.g., object recognition, vital signal measurement, human activity monitoring, object classification / identification, and simultaneous mapping and localization

[0023] , [1], [2]). (ii) MetaWave simulator can be utilized for advancing a general and scalable RF sensing framework and for attacking other frequencies (e.g., Sub-6, 60 / 77-81 GHz, Terahertz), since the physical laws utilized can also be applied to these frequency bands as long as the size of the target is not smaller than the sensing signal wavelength

[0049] , (iii) Besides, the MetaWave attack framework can promote exploring physical attacks on other sensing modalities (e.g., Lidar and infrared camera) in a more comprehensive and explainable way.

[0215] Material Healthiness: Material healthiness is considered in the attack. In order to ensure the stealthiness and practicality of the attack, the attacker needs to wear the material in a regular suit without causing any alert or notice. If the attacker needs extra protection against the harmful material, it will significantly affect stealthiness and fail the attack.

[0216] Attack on mmWave Sensing: As shown in Table III, existing mmWave sensing system attacks utilizing professional electronic transmission devices to spoof location / velocity measurement

[0080] ,

[0060] ,

[0050] ,

[0061] , however, such systems are high cost and can be easily traced by RF sensors on-site. Unlike the previous approaches, MetaWave proposes and designs a new attack type towards mmWave sensing based on low-cost meta-material tags and the simulator. It is also the first work of this attack type.

[0217] Physical Attacks on Sensing'. Existing physical attack methods can be mainly summarized into three categories based on their sensing modality: (1) Camera: Many attacks on cameras use physical patterns to spoof recognition or classification, such as special stickers

[0040] ,

[0035] ,

[0078] , T-shirts

[0089] , or posters

[0087] , (2) Lidar: Most approaches are achieved by placing objects with special geometric shapes

[0072] ,

[0028] ,

[0027] ,

[0081] ,

[0066] to spoof the segmentation or recognition models. (3)

[0218] Microphone'. Most approaches use speakers to play unusual noises over-the-air to make the Automatic Speech Recognition unable to distinguish or misunderstand voice commands

[0071] ,

[0030] ,

[0088] ,

[0025] , Most of these works weigh on the shortcoming of the computing modules. Oppositely, in MetaWave, the fundamental characteristics of mmWave signals are explored and a practical attack approach on the signal side is presented.

[0219] mmWave Sensing with Meta-material Tags'. mmWave sensing has been growing popularity in a variety of domains, such as hand gesture monitoring

[0054] , human activity

[0090] , vital sign monitoring

[0055] ,

[0058] , vibration measurement

[0045] , material types and object status

[0094] ,

[0029] ,

[0058] ,

[0068] ,

[0053] , localization and mapping

[0057] ,

[0084] , and object imaging

[0039] ,

[0095] ,

[0070] , Besides, mmWave sensing is also applied to interact with metamaterial tags enabling new paradigms and applications, such as tagging infrastructure

[0052] , wireless temperature monitoring

[0029] , and through-wall communication

[0032] ,

[0031] , To the inventors’ knowledge, MetaWave is the first work to utilize the metamaterial- enhanced tags to attack mmWave sensing passively.

[0220] A passive attack on mmWave sensing with meta-material -enhanced tags was designed and implemented. The mmWave signal’s properties and passive modulation capabilities are used on RF signals of different metamaterial tags. Then, to optimize the attack, a general attack framework is proposed that includes a simulator of RF and methods that utilize gradient descent to find the optimal parameters of the MetaWave tag. Furthermore, extensive real-world experiments indicated that the proposed MetaWave can achieve an average attack success rate of 97% on range measurement, 96% on angle measurement, and 91% on speed measurement.Various levels of evaluation described herein demonstrate the stealthiness, robustness, and reliability of the proposed system in actual practice. MetaWave has the potential to bring a newresearch vision about the meta-material for wireless sensing and cyberinfrastructure security in the 5G / 6G era.Configuration of Certain Implementations

[0221] The construction and arrangement of the systems and methods as shown in the various implementations are illustrative only. Although only a few implementations have been described in detail in this disclosure, many modifications are possible (e.g., variations in sizes, dimensions, structures, shapes and proportions of the various elements, values of parameters, mounting arrangements, use of materials, colors, orientations, etc.). For example, the position of elements may be reversed or otherwise varied, and the nature or number of discrete elements or positions may be altered or varied. Accordingly, all such modifications are intended to be included within the scope of the present disclosure. The order or sequence of any process or method steps may be varied or re-sequenced according to alternative implementations. Other substitutions, modifications, changes, and omissions may be made in the design, operating conditions, and arrangement of the implementations without departing from the scope of the present disclosure.

[0222] The present disclosure contemplates methods, systems, and program products on any machine-readable media for accomplishing various operations. The implementations of the present disclosure may be implemented using existing computer processors, or by a special purpose computer processor for an appropriate system, incorporated for this or another purpose, or by a hardwired system. Implementations within the scope of the present disclosure include program products including machine-readable media for carrying or having machine-executable instructions or data structures stored thereon. Such machine-readable media can be any available media that can be accessed by a general purpose or special purpose computer or other machine with a processor. By way of example, such machine-readable media can comprise RAM, ROM, EPROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to carry or store desired program code in the form of machine-executable instructions or data structures, and which can be accessed by a general purpose or special purpose computer or other machine with a processor.

[0223] When information is transferred or provided over a network or another communications connection (either hardwired, wireless, or a combination of hardwired or wireless) to a machine, the machine properly views the connection as a machine-readable medium. Thus, any such connection is properly termed a machine-readable medium. Combinations of the above are also included within the scope of machine-readable media. Machine-executable instructions include, for example, instructions and data which cause a general -purpose computer, special purposecomputer, or special purpose processing machines to perform a certain function or group of functions.

[0224] Although the figures show a specific order of method steps, the order of the steps may differ from what is depicted. Also, two or more steps may be performed concurrently or with partial concurrence. Such variation will depend on the software and hardware systems chosen and on designer choice. All such variations are within the scope of the disclosure. Likewise, software implementations could be accomplished with standard programming techniques with rule-based logic and other logic to accomplish the various connection steps, processing steps, comparison steps and decision steps.

[0225] It is to be understood that the methods and systems are not limited to specific synthetic methods, specific components, or to particular compositions. It is also to be understood that the terminology used herein is for the purpose of describing particular implementations only and is not intended to be limiting.

[0226] As used in the specification and the appended claims, the singular forms “a,” “an” and “the” include plural referents unless the context clearly dictates otherwise. Ranges may be expressed herein as from “about” one particular value, and / or to “about” another particular value. When such a range is expressed, another implementation includes-1from the one particular value and / or to the other particular value. Similarly, when values are expressed as approximations, by use of the antecedent “about,” it will be understood that the particular value forms another implementation. It will be further understood that the endpoints of each of the ranges are significant both in relation to the other endpoint, and independently of the other endpoint.

[0227] “Optional” or “optionally” means that the subsequently described event or circumstance may or may not occur, and that the description includes instances where said event or circumstance occurs and instances where it does not.

[0228] Throughout the description and claims of this specification, the word “comprise” and variations of the word, such as “comprising” and “comprises,” means “including but not limited to,” and is not intended to exclude, for example, other additives, components, integers or steps. “Exemplary” means “an example of’ and is not intended to convey an indication of a preferred or ideal implementation. “Such as” is not used in a restrictive sense, but for explanatory purposes.

[0229] Disclosed are components that can be used to perform the disclosed methods and systems. These and other components are disclosed herein, and it is understood that when combinations, subsets, interactions, groups, etc., of these components are disclosed that whilespecific reference of each various individual and collective combinations and permutation of these may not be explicitly disclosed, each is specifically contemplated and described herein, for all methods and systems. This applies to all aspects of this application including, but not limited to, steps in disclosed methods. Thus, if there are a variety of additional steps that can be performed it is understood that each of these additional steps can be performed with any specific implementation or combination of implementations of the disclosed methods.

[0230] The following patents, applications and publications as listed below and throughout this document are hereby incorporated by reference in their entireties herein.REFERENCES[1] “Ai sensing.” [Online], Available: https: / / matrixspace.com / ai-sensing / [2] “Amazon astro.” [Online], Available: https: / / www.amazon.com / dp / B078NSDFSB[3] “Constant false alarm rate (cfar) detection.” [Online],Available:https: / / www.mathworks.com / help / phased / ug / constant-false-alarm-rate-cfar- detection.html[4] “Emf radiation detector.” [Online], Available: https: / / apps.apple.com / us / app / emf-radiation- detector / id 1046907307[5] “Kia safety technology.” [Online], Available: https: / / www.kia.com / us / en / adas[6] “Psrs perimeter surveillance radar system.” [Online], Available: https: / / www.security.honeywell.com / product-repository / psrs-perimeter-surveillance-radar- system[7] “Shenzhen security exhibition: radar security application of the new engine.” [Online],Available: http: / / en.nanoradar.cn / Article / detail / id / 409.html[8] “What is honda sensingR suite? features amp; more: Honda.” [Online], Available: https: / / automobiles.honda.com / sensing[9] “Continental automotive - radars,” April 2022. [Online], Available: https: / / www.continental automotive. com / en-gl / Passenger-Cars / Autonomous-Mobility ZEnablers / Radars

[0010] “Honda civic teardown: Adas (advanced driver-assistance system),” Jan 2022. [Online],Available: https: / / www.marklines.com / en / report all / repl740 201808

[0011] “How speed affects braking distance,” April 2022. [Online], Available: https: / / www.aceable.com / safe-driving / how-speed-affects-braking-distance /

[0012] “Selection of pcb materials for 5g,” April 2022. [Online], Available: https: / / rogerscorp.eom / - / media / proj ect / rogerscorp / documents / articles / english / advanced- connectivity-solutions / selection-of-pcb-materi als-for-5g.pdf

[0013] “A tesla vehicle using ‘smart summon’ appears to crash into a $3.5 million private jet,”April 2022. [Online], Available: https: / / www.theverge.com / 2022 / 4 / 22 / 23037654 / tesla-crash- private-jet-reddit-video-smart-summon

[0014] ““front-radar-sensor.” bosch mobility solutions,” April 2022. [Online], Available: https: / / www.bosch-mobility-solutions.com / en / solutions / sensors / front-radar-sensor /

[0015] “The basics of polarization animated guides,” 2022-04-15. [Online], Available: https: / / www.specac.com / en / news / calendar / 2018 / 04 / polarization-basics

[0016] “C-ram lf-72,” 2022-04-15. [Online], Available: https: / / stores.cumingmicrowave-online- store.com / c-ram-lf-72 /

[0017] “Simulation for connectivity, compatibility, and radar altair feko, ”2022-04-15. [Online],Available: https: / / www.altair.com / feko /

[0018] “Tesla autopilot crashes and causes,” 2022-04-15. [Online], Available: https: / / www.autopilotreview.com / tesla-autopilot-accidents-causes

[0019] “Tesla drivers report a surge in phantom braking,” 2022-04-15. [Online], Available: https: / / www.washingtonpost.com / technology / 2022 / 02 / 02 / tesla-phantom-braking

[0020] R. Appleby and R. N. Anderton, “Millimeter-wave and submillimeterwave imaging for security and surveillance,” Proceedings of the IEEE, vol. 95, no. 8, pp. 1683-1690,2007.

[0021] J. S. Asvestas, “The physical optics method in electromagnetic scattering,” Journal ofMathematical Physics, vol. 21, no. 2, pp. 290-299, 1980.

[0022] J. Baldauf, S.-W. Lee, L. Lin, S.-K. Jeng, S. M. Scarborough, and C. Yu, “High frequency scattering from trihedral corner reflectors and other benchmark targets: Sbr versus experiment,IEEE Transactions on Antennas and Propagation, vol. 39, no. 9, pp. 1345-1351, 1991.

[0023] D. Bames, M. Gadd, P. Murcutt, P. Newman, and I. Posner, “The oxford radar robotcar dataset: A radar extension to the oxford robotcar dataset,” in 2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020, pp. 6433-6438.

[0024] F. O. Bartell, E. L. Dereniak, and W. L. Wolfe, “The theory and measurement of bidirectional reflectance distribution function (brdf) and bidirectional transmittance distribution function (btdf), ” in Radiation scattering in optical systems, vol. 257. SPIE, 1981, pp. 154-160.

[0025] S. Bhattacharya, D. Manousakas, A. G. C. Ramos, S. I. Venieris, N. D. Lane, and C. Mascolo, “Countering acoustic adversarial attacks in microphone-equipped smart home devices,” Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, vol. 4, no. 2, pp. 1-24, 2020.

[0026] A. Boloor, X. He, C. Gill, Y. Vorobeychik, and X. Zhang, “Simple physical adversarial examples against end-to-end autonomous driving models,” in 2019 IEEE International Conference on Embedded Software and Systems (ICESS). IEEE, 2019, pp. 1-7.

[0027] Y. Cao, N. Wang, C. Xiao, D. Yang, J. Fang, R. Yang, Q. A. Chen, M. Liu, and B. Li, “ Invisible for both camera and lidar: Security of multi-sensor fusion based perception in autonomous driving under physical -world attacks,” in 2021 IEEE Symposium on Security and Privacy (SP). IEEE, 2021, pp. 176-194.

[0028] Y. Cao, C. Xiao, B. Cyr, Y. Zhou, W. Park, S. Rampazzi, Q. A. Chen, K. Fu, and Z. M. Mao, “Adversarial sensor attack on lidar-based perception in autonomous driving,” in Proceedings of the 2019 ACM SIGS AC conference on computer and communications security, 2019, pp. 2267-2281.

[0029] B. Chen, H. Li, Z. Li, X. Chen, C. Xu, and W. Xu, “Thermowave: a new paradigm of wireless passive temperature monitoring via mmwave sensing,” in Proceedings of the 26th Annual International Conference on Mobile Computing and Networking, 2020, pp. 1-14.

[0030] G. Chen, S. Chenb, L. Fan, X. Du, Z. Zhao, F. Song, and Y. Liu, “Who is real bob? adversarial attacks on speaker recognition systems,” in 2021 IEEE Symposium on Security andPrivacy (SP). IEEE, 2021, pp. 694-711.

[0031] L. Chen, W. Hu, K. Jamieson, X. Chen, D. Fang, and J. Gummeson “Pushing the physical limits of iot devices with programmable metasurfaces,” in 18th {USENIX} Symposium on Networked Systems Design and Implementation ({NSDI} 21), 2021, pp. 425-438.

[0032] K. W. Cho, M. H. Mazaheri, J. Gummeson, O. Abari, and K. Jamieson, “mmwall: A reconfigurable metamaterial surface for mmwave networks,” in Proceedings of the 22ndInternational Workshop on Mobile Computing Systems and Applications, 2021, pp. 119-125

[0033] M. Di Renzo, K. Ntontin, J. Song, F. H. Danufane, X. Qian, F. Lazarakis, J. De Rosny, D - T. Phan-Huy, O. Simeone, R. Zhang et al., “Reconfigurable intelligent surfaces vs. relaying:Differences, similarities, and performance comparison,” IEEE Open Journal of the Communications Society, vol. 1, pp. 798-807, 2020.

[0034] DJI, “Dji mini se,” 2021. [Online], Available: https: / / www.dji.com / mini-se

[0035] I. Evtimov, K. Eykholt, E. Fernandes, T. Kohno, B. Li, A. Prakash, A. Rahmati, and D. Song, “Robust physical -world attacks on machine learning models,” arXiv preprint arXiv: 1707.08945, vol. 2, no. 3, p. 4, 2017.

[0036] FAA, “How to register your drone,” 2021. [Online], Available: https: / / www.faa.gov / uas / getting$\ $started / register$\ Sdrone /

[0037] W. Gao, M. Sui, and X. Xu, “Impact of the divergence factor on accurate res calculation using shooting and bouncing rays,” in 2011 4thIEEE International Symposium on Microwave, Antenna, Propagation and EMC Technologies for Wireless Communications. IEEE, 2011, pp. 206-209.

[0038] W. C. Gibson, The method of moments in electromagnetics. Chapman and Hall / CRC, 2007.

[0039] J. Guan, S. Madani, S. Jog, S. Gupta, and H. Hassanieh, “Through fog high-resolution imaging using millimeter wave radar,” in Proceedings of the IEEE / CVF Conference on Computer Vision and Pattern Recognition, 2020.

[0040] Y. Guo, X. Wei, G. Wang, and B. Zhang, “Meaningful adversarial stickers for face recognition in physical world,” arXiv preprint arXiv:2104.06728, 2021.

[0041] M. Z. Ikram, A. Ahmad, and D. Wang, “High-accuracy distance measurement using millimeter-wave radar,” in 2018 IEEE Radar Conference (RadarConfl8). IEEE, 2018, pp. 1296 1300.

[0042] A. Ilyas, L. Engstrom, A. Athalye, and J. Lin, “Black-box adversarial attacks with limited queries and information,” in International Conference on Machine Learning. PMLR, 2018, pp. 2137-2146.

[0043] Infineon, “24ghz transceiver,” 2021. [Online], Available: https: / / www.infmeon.com / dgdl / Infmeon-AN553 $\ $BGT24MTR12$\ $XMC4700$\ $Position2Go$\ $DemoBoard-ApplicationNotes-v01 $\ $02EN.pdf?fileId=5546d4626cb27db2016d44631 adb021 f

[0044] - , “Demo position2go,” 2021. [Online], Available: https: / / www.infmeon.com / cms / en / product / evaluation-boards / demo-position2go /

[0045] C. Jiang, J. Guo, Y. He, M. Jin, S. Li, and Y. Liu, “mmvib: micrometerlevel vibration measurement with mmwave radar,” in Proceedings of the 26th Annual International Conference on Mobile Computing and Networking, 2020, pp. 1-13.

[0046] J.-M. Jin, The finite element method in electromagnetics. John Wiley & Sons, 2015.

[0047] C. R. Karanam, B. Korany, and Y. Mostofi, “Magnitude-based angleof-arrival estimation, localization, and target tracking,” in 2018 17thACM / IEEE International Conference on Information Processing in Sensor Networks (IPSN). IEEE, 2018, pp. 254-265

[0048] J. Kiefer and J. Wolfowitz, “Stochastic estimation of the maximum of a regression function, ” The Annals of Mathematical Statistics, pp. 462-466, 1952.

[0049] R. E. Kleinman, “The rayleigh region,” Proceedings of the IEEE, vol. 53, no. 8, pp. 848- 856, 1965.

[0050] R. Komissarov and A. Wool, “Spoofing attacks against vehicular fmcw radar,” arXiv preprint arXiv:2104.13318, 2021.

[0051] T.-M. Li, M. Aittala, F. Durand, and J. Lehtinen, “Differentiable monte carlo ray tracing through edge sampling,” ACM Transactions on Graphics (TOG), vol. 37, no. 6, pp. 1-11, 2018.

[0052] Z. Li, B. Chen, Z. Yang, H. Li, C. Xu, X. Chen, K. Wang, and W. Xu, “Ferrotag: A paperbased mmwave-scannable tagging infrastructure,” in Proceedings of the 17th Conference on Embedded Networked Sensor Systems, 2019, pp. 324-337.

[0053] Z. Li, F. Ma, A. S. Rathore, Z. Yang, B. Chen, L. Su, and W. Xu, “Wavespy: Remote and through-wall screen attack via mmwave sensing,” in 2020 IEEE Symposium on Security and Privacy (SP). IEEE, 2020, pp. 217-232.

[0054] J. Lien, N. Gillian, M. E. Karagozler, P. Amihood, C. Schwesig, E. Olson, H. Raja, and I. Poupyrev, “Soli: Ubiquitous gesture sensing with millimeter wave radar,” ACM Transactions on Graphics (TOG), vol. 35, no. 4, pp. 1-19, 2016.

[0055] F. Lin, C. Song, Y. Zhuang, W. Xu, C. Li, and K. Ren, “Cardiac scan: A non-contact and continuous heart-based user authentication system,” in Proceedings of the 23rd Annual International Conference on Mobile Computing and Networking, ser. MobiCom ’17. New York, NY, USA: ACM, 2017, pp. 315-328.

[0056] H. Ling, R.-C. Chou, and S.-W. Lee, “Shooting and bouncing rays: Calculating the res of an arbitrarily shaped cavity,” IEEE Transactions on Antennas and propagation, vol. 37, no. 2, pp. 194-205, 1989.

[0057] C. X. Lu, S. Rosa, P. Zhao, B. Wang, C. Chen, J. A. Stankovic, N. Trigoni, and A. Markham, “See through smoke: robust indoor mapping with low-cost mmwave radar,” in Proceedings of the 18th International Conference on Mobile Systems, Applications, and Services, 2020, pp. 14-27.

[0058] C. X. Lu, M. R. U. Saputra, P. Zhao, Y. Almalioglu, P. P. de Gusmao, C. Chen, K. Sun, N. Trigoni, and A. Markham, “milliego: single-chip mmwave radar aided egomotion estimation via deep sensor fusion,” in Proceedings of the 18th Conference on Embedded Networked Sensor Systems (SenSys), 2020.

[0059] T. M' oller and B. Trumbore, “Fast, minimum storage ray-triangle intersection,” Journal of graphics tools, vol. 2, no. 1, pp. 21-28, 1997.

[0060] P. Nallabolu and C. Li, “A frequency-domain spoofing attack on fmcw radars and its mitigation technique based on a hybrid-chirp waveform,” IEEE Transactions on Microwave Theory and Techniques, vol. 69, no. 11, pp. 5086-5098, 2021.

[0061] S. Nashimoto, D. Suzuki, N. Miura, T. Machida, K. Matsuda, and M. Nagata, “Low-cost distance-spoofing attack on fmcw radar and its feasibility study on countermeasure,” Journal of Cryptographic Engineering, pp. 1-10, 2021.

[0062] P. Nguyen, V. Kakaraparthi, N. Bui, N. Umamahesh, N. Pham, H. Truong, Y. Guddeti, D. Bharadia, R. Han, E. Frew et al., “Dronescale: drone load estimation via remote passive rf sensing,” in Proceedings of the 18th Conference on Embedded Networked Sensor Systems,2020, pp. 326-339.

[0063] R. Nitzberg, “Constant-false-alarm-rate signal processors for several types of interference,” IEEE Transactions on Aerospace and Electronic Systems, no. 1, pp. 27-34, 1972.

[0064] J. Nolan, K. Qian, and X. Zhang, “Ros: passive smart surface for roadside-to-vehicle communication,” in Proceedings of the 2021 ACM SIGCOMM 2021 Conference, 2021, pp.165- 178.

[0065] H. W. Ott and H. W. Ott, Noise reduction techniques in electronic systems. Wiley New York, 1988, vol. 442.

[0066] W. Park, N. Liu, Q. A. Chen, and Z. M. Mao, “Sensor adversarial traits: Analyzing robustness of 3d object detection sensor fusion models,” in 2021 IEEE International Conference on Image Processing (ICIP). IEEE, 2021, pp. 484-488.

[0067] C. Peterson, “How police radar works, how to avoid it with the best radar detector,” Jan2021. [Online], Available: https: / / radartest.com / how-radar-works.asp

[0068] A. Prabhakara, V. Singh, S. Kumar, and A. Rowe, “Osprey: a mmwave approach to tire wear sensing,” in Proceedings of the 18th International Conference on Mobile Systems, Applications, and Services, 2020.

[0069] K. Qian, Z. He, and X. Zhang, “3d point cloud generation with millimeter-wave radar,” Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, vol. 4, no. 4, pp. 1-23, 2020.

[0070] K. Qian, S. Zhu, X. Zhang, and L. E. Li, “Robust multimodal vehicle detection in foggy weather using complementary lidar and radar signals,” in Proceedings of the IEEE / CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 444-453.

[0071] Y. Qin, N. Carlini, G. Cottrell, I. Goodfellow, and C. Rafifel, “Imperceptible, robust, and targeted adversarial examples for automatic speech recognition,” in International conference on machine learning. PMLR, 2019, pp. 5231-5240.

[0072] H. Ren and T. Huang, “Adversarial example attacks in the physical world,” in International Conference on Machine Learning for Cyber Security. Springer, 2020, pp. 572-582.

[0073] M. A. Richards, Fundamentals of radar signal processing. McGraw-Hill Education, 2014.

[0074] H. Robbins and S. Monro, “A stochastic approximation method,” The annals of mathematical statistics, pp. 400-407, 1951.

[0075] D. R. I. M. Setiadi, “Psnr vs ssim: imperceptibility quality assessment for image steganography,” Multimedia Tools and Applications, vol. 80, pp. 8423-8444, 2021.

[0076] M. Shahid, A. Rossholm, B. L' ovstr' om, and H.-J. Zepernick, “Noreference image and video quality assessment: a classification and review of recent approaches,” EURASIP Journal on image and Video Processing, vol. 2014, no. 1, pp. 1-32, 2014.

[0077] J. Shen, N. Wang, Z. Wan, Y. Luo, T. Sato, Z. Hu, X. Zhang, S. Guo, Z. Zhong, K. Li et al., “Sok: On the semantic ai security in autonomous driving,” arXiv preprint arXiv:2203.05314, 2022

[0078] D. Song, K. Eykholt, I. Evtimov, E. Fernandes, B. Li, A. Rahmati, F. Tramer, A.Prakash, and T. Kohno, “Physical adversarial examples for object detectors,” in 12th {USENIX} Workshop on Offensive Technologies ({WOOT} 18), 2018.

[0079] P. Staat, H. Elders-Boll, M. Heinrichs, C. Zenger, and C. Paar, “Mirror mirror on the wall: Wireless environment reconfiguration attacks based on fast software-controlled surfaces,” arXiv preprint arXiv:2107.01709, 2021.

[0080] Z. Sun, S. Balakrishnan, L. Su, A. Bhuyan, P. Wang, and C. Qiao, “Who is in control? practical physical layer attack and defense for mmwave-based sensing in autonomous vehicles,” IEEE Transactions on Information Forensics and Security, vol. 16, pp. 3199-3214, 2021.

[0081] J. Tu, M. Ren, S. Manivasagam, M. Liang, B. Yang, R. Du, F. Cheng, and R. Urtasun, “ Physically realizable adversarial examples for lidar object detection,” in Proceedings of the IEEE / CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 13 716-13 725.

[0082] M. Vinkler, J. Bittner, and V. Havran, “Extended morton codes for high performance bounding volume hierarchy construction,” in Proceedings of High Performance Graphics, 2017, pp. 1-8.

[0083] Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, “Image quality assessment: from error visibility to structural similarity,” IEEE transactions on image processing, vol. 13, no. 4, pp. 600-612, 2004.

[0084] T. Wei and X. Zhang, “mtrack: High-precision passive tracking using millimeter wave radios,” in Proceedings of the 21st Annual International Conference on Mobile Computing and Networking, 2015, pp. 117-129.

[0085] Z. Wei, F. Zhang, S. Chang, Y. Liu, H. Wu, and Z. Feng, “Mmwave radar and vision fusion for object detection in autonomous driving: A review,” Sensors, vol. 22, no. 7, p. 2542, 2022.

[0086] V. Winkler, “Range doppler detection for automotive fmcw radars,” in 2007 European Radar Conference. IEEE, 2007, pp. 166-169.

[0087] R. R. Wiyatno and A. Xu, “Physical adversarial textures that fool visual object tracking,” in Proceedings of the IEEE / CVF International Conference on Computer Vision, 2019, pp. 4822- 4831.

[0088] Y. Xie, C. Shi, Z. Li, J. Liu, Y. Chen, and B. Yuan, “Real-time, universal, and robust adversarial attacks against speaker recognition systems,” in ICASSP 2020-2020 IEEE international conference on acoustics, speech and signal processing (ICASSP). IEEE, 2020, pp. 1738-1742.

[0089] K. Xu, G. Zhang, S. Liu, Q. Fan, M. Sun, H. Chen, P.-Y. Chen, Y. Wang, and X. Lin,Adversarial t-shirt! evading person detectors in a physical world,” in European Conference on Computer Vision. Springer, 2020, pp. 665-681.

[0090] H. Xue, Y. Ju, C. Miao, Y. Wang, S. Wang, A. Zhang, and L. Su, “mmmesh: towards 3d real-time dynamic human mesh construction using millimeter- wave,” in Proceedings of the 19thAnnual International Conference on Mobile Systems, Applications, and Services, 2021, pp. 269- 282.

[0091] H. Yang, X. Cao, F. Yang, J. Gao, S. Xu, M. Li, X. Chen, Y. Zhao, Y. Zheng, and S. Li, “A programmable metasurface with dynamic polarization, scattering and focusing control,” Scientific reports, vol. 6, no. 1, pp. 1-11, 2016.

[0092] P. Zhao, C. X. Lu, B. Wang, C. Chen, L. Xie, M. Wang, N. Trigoni, and A. Markham, “Heart rate sensing with a robot mounted mmwave radar,” in 2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020, pp. 2812-2818.

[0093] Y. Zheng, Y. Zhou, J. Gao, X. Cao, H. Yang, S. Li, L. Xu, J. Lan, and L. Jidi, “Ultra wideband polarization conversion metasurface and its application cases for antenna radiation enhancement and scattering suppression,” Scientific reports, vol. 7, no. 1, pp. 1-12, 2017.

[0094] Y. Zhu, Y. Zhu, B. Y. Zhao, and H. Zheng, “Reusing 60ghz radios for mobile radar imaging,” in Proceedings of the 21st Annual International Conference on Mobile Computing and Networking, 2015, pp. 103-116.

[0095] Y. Zhu, Y. Zhu, Z. Zhang, B. Y. Zhao, and H. Zheng, “60ghz mobile imaging radar,” in Proceedings of the 16th International Workshop on Mobile Computing Systems and Applications, 2015, pp. 75-80.

Claims

WHAT IS CLAIMED IS:

1. A system comprising: a processor; and memory having instructions stored thereon that, when executed by the processor, cause the system to: receive target parameters and scene parameters for a potential millimeter-wave (mm Wave) attack on a target entity; generate a set of initial parameters for a passive mmWave attack tag, wherein the mmWave attack tag is configured to disrupt radiofrequency (RF) signals emitted by the target entity; automatically determine an optimized set of parameters for the passive mmWave attack tag by iteratively modifying the set of initial parameters for the passive mmWave attack tag until a calculated attack loss is less than a predefined threshold and / or a maximum number of iterations is reached; and output, to a user of the system, the optimized set of parameters for the passive mmWave attack tag, wherein the passive mmWave attack tag is used to conduct tests with respect to the target entity under real-world conditions.

2. The system of claim 1, wherein the passive mmWave attack tag is at least partially formed from one or more of Counter Rocket, Artillery, and Mortar - Lightweight System (C- RAM LF), tin foil, or a copper parallel wire grid.

3. The system of claim 1 or 2, wherein the passive mmWave attack tag is a malleable two- dimensional or three-dimensional structure.

4. The system of any one of claims 1-3, wherein the passive mmWave attack tag comprises at least one of a hexagonal shape or honeycomb structure.

5. The system of any one of claims 1-4, wherein the passive mmWave attack tag is configured to adhere to a surface of at least one of a vehicle, autonomous robot, or stationary structure during the tests.

6. The system of any one of claims 1-5, wherein the passive mmWave attack tag is configured to disrupt the RF signals emitted by the target entity by one of absorbing at least aportion of the RF signals, reflecting a majority of the RF signals back to the target entity, or modifying a polarization of the RF signals.

7. The system of any one of claims 1-6, wherein the target parameters comprise at least one of a frequency, bandwidth, polarization mode, or sensing distance of a radar sensor operated by the target entity.

8. The system of any one of claims 1-7, wherein the scene parameters comprise a geometry and / or material properties of at least one object within a mmWave attack environment.

9. The system of any one of claims 1-8, wherein the set of initial parameters comprises at least one of a material, a pattern, a size, a position, and a rotation of the passive mmWave attack tag.

10. The system of any one of claims 1-8, wherein the RF signals emitted by the target entity are between 24 GHz and 300 GHz.

11. A method comprising: receiving target parameters and scene parameters for a potential millimeter-wave (mmWave) attack on a target entity; generating a set of initial parameters for a passive mmWave attack tag, wherein the passive mmWave attack tag is configured to disrupt radiofrequency (RF) signals emitted by the target entity; automatically determining an optimized set of parameters for the passive mmWave attack tag by iteratively modifying the set of initial parameters for the passive mmWave attack tag until a calculated attack loss is less than a predefined threshold and / or a maximum number of iterations is reached; and outputting, to a user, the optimized set of parameters for the passive mmWave attack tag, wherein the passive mmWave attack tag is used to conduct tests with respect to the target entity under real-world conditions.

12. The method of claim 11, wherein the passive mmWave attack tag is at least partially formed from one or more of C-RAM LF, tin foil, or a copper parallel wire grid.

13. The method of claim 11 or 12, wherein the passive mmWave attack tag comprises a malleable two-dimensional or three-dimensional structure.

14. The method of any one of claims 11-13, wherein the passive mmWave attack tag is configured to disrupting the RF signals emitted by the target entity by one of absorbing at least a portion of the RF signals, reflecting a majority of the RF signals back to the target entity, or modifying a polarization of the RF signals.

15. The method of any one of claims 11-14, wherein the target parameters comprise at least one of a frequency, bandwidth, polarization mode, or sensing distance of a radar sensor operated by the target entity.

16. The method of any one of claims 11-15, wherein the scene parameters comprise a geometry and material properties of at least one object within a mmWave attack environment.

17. The method of any one of claims 11-16, wherein the set of initial parameters comprises at least one of a material, a pattern, a size, a position, and a rotation of the passive mmWave attack tag.

18. The method of any one of claims 11-17, wherein the RF signals emitted by the target entity are between 24 GHz and 300 GHz.

19. A method comprising: obtaining target parameters and scene parameters for a potential millimeter-wave (mmWave) attack on a target entity; obtaining parameters for a passive mmWave attack tag, wherein the passive mmWave attack tag is configured to disrupt radiofrequency (RF) signals emitted by the target entity; generating a digital attack environment for the mmWave attack based on the target parameters and the scene parameters; simulating emission of the RF signals emitted by the target entity within the digital attack environment; calculating reflected RF signals from the passive mmWave attack tag based on the parameters of the passive mmWave attack tag; determining simulated time-domain RF signals from the reflected RF signals; andusing the simulated time-domain RF signals to optimize the parameters of the passive mm Wave attack tag.

20. The method of claim 19, wherein the target parameters comprise at least one of a frequency, bandwidth, polarization mode, or sensing distance of a radar sensor operated by the target entity.

21. The method of claim 19 or 20, wherein the scene parameters comprise a geometry and material properties of at least one object within a mmWave attack environment.

22. The method of any one of claims 19-21, wherein the reflected RF signals are calculated using ray tracing.

23. The passive mmWave attack tag according to any of the system of claims 1-10 or the methods of claims 11-22.

Citation Information

Patent Citations

  • Millimeter-wave radar imaging device and method

    US20210255313A1

  • Portable handheld security device

    WO2007008217A1