TACTICAL CROSS-DOMAIN SOLUTIONS: CURRENT STATUS AND THE NEED FOR CHANGE Kevin Plyler General Dynamics C4 Systems kevin.plyler@gdc4s.com Brian C. Tague NGC-TASC brian.tague@ngc.com Dr. Roshan Thomas Cobham Analytic Services roshan.thomas@cobham.com ABSTRACT l The rapid migration of system high 2 information sharing to the tactical edge has made it imperative that the DoD reexamine tactical Cross Domain Solutions/Enterprise Services (CDS/ES) 3. Prior to Operation Iraqi Freedom (OIF), information sharing requirements at the tactical edge were relatively few in number and nominal in terms of data throughput, data types, and users. Cross Domain Solutions (CDS)4 deployed back then were specialized, hardened, and resistant to hacking in the event of enemy overrun. Since OIF, both the volume ofbattle field system high tactical networks 5 as well as the operational requirements to support each of these networks (i.e., increased data throughput, data types and variety ofusers) have significantly increased [On Point]. When combined with additional constraints inherent to the battle space such as low latency, Size, Weight and Power (SWaP), the current approach to addressing information sharing requirements in a tactical network breaks down. Taking the traditional, point-to-point approach by making a CDS smaller and more robust in the tactical environment may be adequate in the near term. However, this approach will not support the requirements levied upon next generation warfighting systems. In the future, interdependent tactical Prepared through collaborative participation in the Communications and Networks Consortium sponsored by the U. S. Army Research Laboratory under the Collaborative Technology Alliance Program, Cooperative Agreement DAAD19-01-2-0011. The U. S. Government is authorized to reproduce and distribute reprints for Government purposes notwithstanding any copyright notation thereon. 2 The highest classification level of the data on an Information Sharing system. 3 Term referring to either a point-to-point Cross Domain Solution (CDS) or Cross Domain Enterprise Service (CDES). Each of these is further defmed in the paper. 4 A form of controlled interface that provides the ability to manually and/or automatically access and/or transfer information between different security domains. 5 Tactical Networks: Mobile ad hoc networks that can operate with full mobility, self-configuration, robustness, survivability, and scalability. 978-1-4244-5239-2/09/$26.00 ©2009 IEEE Dr. Simon Tsang Telcordia stsang@telcordia.com networks will be required to exhibit a dynamic (self organizing) nature, supporting adaptability and quick response to data ingress and egress. Nodes on these future networks will also need to operate with severely limited bandwidth and other operational and/or environment constraints. Therefore it is necessary to examine current and future information sharing requirements at the tactical edge from both the CDS/ES developer as well as user perspective. This position paper will discuss both perspectives in order to allow a better understanding ofthe current CD problem space, as well as gain insight into building the next generation CDS/ES. 1 INTRODUCTION In the accelerated pace of combat operations since Operation Iraqi Freedom (OIF), the ability to securely move data between system high networks via CDS/ES has been outstripped by the deployment of tactical networks to the battle space. During the period leading up to OIF, system high information sharing systems deployed at the tactical edge were relatively limited in terms of data throughput, fixed data types, number of users, and information assurance risk exposure. CDS/ES consisted of installing and supporting a finite number of guards" that were specialized, hardened, and resistant to information assurance threats. The fundamental security risk was enemy overrun. Since OIF, deployments of system high tactical networks in the battle space have proliferated along with the digitization' of information by the armed services [On Point]. Current security requirements (Le., JDISS 8 ) call Guard: A device or collection of devices that mediate controlled transfers of information across security boundaries (e.g., between Security Domain "A" and Security Domain "B") 7 Digitization initiatives include Army's Force XXI, the Air Force's Effects-Based Operations, the Navy's Cooperative Engagement, and Marine Corps Sea Dragon. 8 Joint Deployable Intelligence Support System (JDISS) one of several sources of security requirements related to deployable systems provides a family of hardware and software capabilities 6 Page 1 of7 Authorized licensed use limited to: Defence Science and Technology Group (DSTG). Downloaded on January 25,2023 at 00:00:24 UTC from IEEE Xplore. Restrictions apply. for advancing deployment of CDS/ES from operation centers to become integrated into vehicle based, hand held, and other portable devices [On Point]. However, developing CDS/ES to support these requirements at the tactical level (see figure 1) is a difficult problem without a near term solution. Existing Tactical CDS V. - Fixed data formats - Simple filters/hardwired - Centralized security controls - Non configurable Evolving Tactical CDS - Evolving data formats - Complex software filters - Few admin rules - Reduced administrative access - Complex admin rules - Broad administrative access Figure 1-1 Computing) Changing Requirements CROSS-DOMAIN SOLUTIONS (CDS) SOME DEFINITIONS Multiple Security Levels (MSL) - An architecture where system high networks exchange information via a controlled mediating device (e.g., CDS) [MITRE]. Multi-Level Security (MLS) - Concept of processing information with different classifications and categories that simultaneously permits access by users with different security clearances and denies access to users who lack authorization [MITRE]. - Distributed security controls - Configurable CDS/ES 2 (Owl In addition to the risk of enemy overrun, tactical networks also pose additional key security risks with regard to: o Increases in data throughput via multiple networks o Evolving and more complex data formats o Number and variety of users o Multiple opportunities for malicious data injection/attack. The current approach to addressing CDS/ES for information sharing requirements for networks in a nontactical environment breaks down when applied to the tactical environment. The low latency, Size Weight and Power (SWaP) constraints inherent to tactical networks, combined with risk of enemy overrun, further exacerbate the already complex and intricate security challenges posed by these new requirements. It may be feasible to continue to make CDS/ES smaller and more robust in the near term to support the full set of information sharing requirements in a tactical environment. However it would be prohibitively difficult, risky and expensive to deploy and operate thousands of such CDS/ES to support next generation warfighting systems in the long term. In order to share information in a robust and secure fashion at the tactical edge in the near term, CDS/ES configurations must resolve risks under declining SWaP constraints. In order to achieve this, next generation information sharing systems must deliver some type of CD Enterprise Service (CDES) scalable to tactical networks. that allow connectivity and interoperability with intelligence systems supporting forces, in garrison, and deployed. Multilevel Security (MLS) vs Figure 2-1 MLS V. MSL (MITRE) Mandatory Access Control/Discretionary Access Control (MACIDAC) - MAC: Means of restricting access to objects based on the sensitivity of the information contained in the objects and the formal authorization (i.e., clearance, formal access approvals, and need-to-know) of subjects to access information of such sensitivity. MAC is commonly used in MLS to prevent subjects from circumventing data flow policy. DAC: Means of restricting access to objects based on the identity and need-to-know of users and/or groups to which the object belongs. Controls are discretionary in the sense that a subject with certain access permission is capable of passing that permission (directly or indirectly) to any other subject. Cross Domain Access (CDA) These provide simultaneous visual access to multiple security domains via a single workstation (i.e., client/workstations) [CD Inventory]. Cross Domain Transfer (CDT) - These interconnect networks of information systems that operate in different security domains, and mediate transfer of information between them [CD Inventory]. Information is 'transferred' Page 2 of7 Authorized licensed use limited to: Defence Science and Technology Group (DSTG). Downloaded on January 25,2023 at 00:00:24 UTC from IEEE Xplore. Restrictions apply. across domains via accredited transfer devices (Le., "guards"), Cross Domain Multi-level Environment (CDM) - These are used to store data in multiple security domains at varied security levels and allow users to access the data at an appropriate security level [CD Inventory]. CDMs enable access based upon authorization to data from a managed label-aware environment (Le., "servers"). Communities of Interest (COl) - As U.S. joint operations are leveraged, and international cooperation continues to advance, communities of interest continue to form. These communities need to share information in a well controlled fashion while maintaining protection of other information. Examples of these communities are JWICs, TS/SCI, SIPRNet, NIPRNet, TacticalUnclassified, Coalition, NATO, and other ad-hoc relationships. Compounding this problem is that each enclave" to enclave interconnect requires a unique cross domain solution. The result in some cases can be a mesh of networks with a CDS/ES at each point of intersection. As CDS/ES needs move to the battlefield, the number of instances of each of these CDSIES networks expands as well (Le., ship networks, space nets, ground vehicle groups, operation centers, etc) [CJCSI]. If COIs overlap (Le., users are members of multiple COIs), this further complicates the task of CDSIES that are managing information flows as even simple information sharing policies can become difficult to evaluate effectively, and policy enforcement may need to be applied at even greater granularity - to the per user/node level[IAMANET]. 3 CROSS-DOMAIN SOLUTIONS/ENTERPRISE SERVICES (CDS/ES) - STRATEGIC ISSUES 3.1 CDS/ES ARE EXPENSIVE While there is no question that CDS/ES is required for the tactical edge, CDSIES is expensive in terms of both development and deployment costs. Development & Production Costs: Current CDS/ES are custom solutions with limited ability to architect solutions by leveraging common off-the-shelf components. Consequently, such solutions are expensive and slow to develop. Reasons for this include the following: An enclave is a collection of computing environments connected by one or more internal networks under the control of a single authority and security policy, including personnel and physical security - NSTISSI-4009 9 • • • • There is currently a lack of standards for CDSIES, in particular CDSIES architecture and interfaces. This leads to custom solutions that are monolithic or stovepiped solutions that can lead to integration issues. The CDS/ES certification process is cumbersome and slow, increasing overall development costs and leading to production delays. Current solutions must be deployed with pre-defined and certified CDSIES security policies, delaying overall production. There is a lack of standards for CDS/ES policy languages, so policies are custom-developed for particular platforms, and policies developed for one platform are incompatible with other platforms or solutions. This lack of policy re-use leads to increased costs for policy definition and verification. General adoption of existing policy standards such as the IETF Policy Core Information Model [PC1M] will improve the current situation. However, further research is needed to develop policy languages that are tailored for cross-domain security policies [CDL 08]. Deployment Costs: Due to their specialized nature, current CDS/ES incur high deployment costs. These costs must be considered and weighed against other deployment costs. Reasons for high CDS/ES deployment cost include the following. • The current generation of CDSIES requires specialized hardware (primarily to meet assurance requirements) that is several times more expensive than general purpose computing solutions. • CDS/ES that require special hardware are harder to deploy, manage and maintain. • CDSIES is a "special solution" that does not interact easily with common windowing platforms and applications, leading to a lack of integration, administration, and usability. • CDSIES may require modification of existing fielded software or software configurations so that they can be used with the CDSIES. • CDS/ES maintenance requirements are typically more stringent that general computing and single level computing systems. This leads to longer maintenance cycles and additional maintenance costs. 3.2 SECURITY AND ROBUSTNESS IN A TACTICAL ENVIRONMENT As difficult or daunting the notion of developing solid, robust CDS/ES to share information, more recent examples only serve to highlight the problem. These include the misuse of removable media (e.g., thumb Page 3 of7 Authorized licensed use limited to: Defence Science and Technology Group (DSTG). Downloaded on January 25,2023 at 00:00:24 UTC from IEEE Xplore. Restrictions apply. drives), and cross connecting cabling which presents serious security concerns and introduces instability to the attached networks as the connections are added and removed. The scale and scope at which information sharing needs to be accomplished today cannot be sustained by manual (i.e., "sneaker net") methods. When manual methods of information sharing are used in place of electronic methods, oftentimes many critical procedures are bypassed, knowingly or unknowingly, introducing significant, often incalculable risk to the overall mission. This is exacerbated even more in wartime [MITRE]. In the past, cross domain transfers were made using 'air gap' separation between information systems with inexpensive media (e.g., tape, disks, flash drives), and human review as the popular solution. This approach is fraught with opportunities for failure and inconsistent results [MITRE]. The more modern approach has been for information systems to perform automated procedures for mediating upgrades and downgrades of data. These systems mitigated inadvertent disclosure of information and data poisoning issues. The method used is strict adherence to fixed format and known vocabulary data restrictions. These automated transfers are augmented by human review for more complex and subjective data forms (e.g., video, imagery, etc... ). controls designed into the CDSIES system solution to be appropriately mitigated. In addition, a tactical environment typically also has additional constraints on a CDS/ES. A CDS/ES may be in, or near the weapons fuse chain or between a sensor and warfighter, demanding a higher degree of real-time determinism than in an enterprise setting. While there will always be a dynamic dissonance between what is required versus what is available to meet ongoing operational requirements, CDSIES pose particular difficulties for the warfighter who relies heavily upon secure, timely and available information from tactical networks. Several causes for this lag include: • The level of complexity involved with developing a 'trusted product' that supports the warfighter with both rapidly evolving capabilities as well as a high level of assurance [MITRE]. • The relatively few 'trusted products' that are currently available to meet these evolving warfighter requirements [MITRE]. • The lack of commercial incentives and infrastructure to balance the desire to share with the need to protect the information that is shared [MITRE]. 3.3 WARFIGHTER NEEDS The common and traditional location of a CDSIES is in an enterprise setting. Weapon system laboratories, research facilities, joint operations centers, regional operations offices, and academic facilities are examples of these environments. These locations are physically secure, but are none-the-Iess still at risk from information disclosure by authorized and unauthorized personnel either intentionally or inadvertently. Mis-configuration, traffic analysis, malicious code, inappropriate use and inherent information systems vulnerabilities are all risks that must be factored into the use of all CDSIES. Coping with high stress and cognitive load. The warfighter is typically in a high stress environment dealing with very difficult and rapidly evolving task orders. This may involve multitasking and imposes high cognitive load. Accordingly, the environment has to provide multiple modalities in terms of input and output devices. He may thus be interacting with the CDS/ES wearing environmental gloves, helmet-mounted displays all while bouncing in a moving vehicle and dealing with the constraints of a recent affliction or impairment. The war fighter may commonly be in a hurry due to the timely need of such data. Tactical environments present additional concerns to the application of a CDSIES. In a tactical setting, a CDSIES is more likely to be exposed to the physical rigors of the environment. Increased risks of access by unauthorized personnel and a realistic possibility of overrun by an adversary are genuine concerns. A tactical environment also exposes the attached networks and infrastructure to attack. There is also a legitimate risk that these assets, both hardware and software (e.g., security and flow policy) could be subverted in the field, or even removed and taken back to an adversary's lab for further and detailed exploitation. All of these concerns need to have proper Intuitive and safe user interface design. CDS/ES need to accommodate the dynamics of the above conditions and provide a human level interface that meets several criteria. The interface needs to be clear and certain and reduce the overall cognitive load. User interface management systems that factor in psycho-physiological input may be applicable here. Such a system could monitor stress level, heart rate, error rate, etc, and adapt the data presentation, data input, data output, and menu navigation techniques. A war fighter performing human review needs to clearly see what data is being re-graded, what the source and Page 4 of7 Authorized licensed use limited to: Defence Science and Technology Group (DSTG). Downloaded on January 25,2023 at 00:00:24 UTC from IEEE Xplore. Restrictions apply. destinations are, and any analysis the CDS/ES has performed (e.g., dirty word identification), with a simple and clear indication stating that the re-grade request is approved/disapproved. Enhance data security and reduce the risks of data leakage. The computing and user interface environment must enhance data safety and reduce risks of accidental or malicious leakage. Data and indicator markings need to be large and obvious so that decisions are made intentionally and the results are well understood. The design challenge here is to provide the correct tradeoffs between ease of use and data security. Lastly, the data must remain safe. For example, ease of use may be enhanced by oversimplifying the interface to provide cut/copy/paste type features between domains. However such an action should not compromise the sender and recipient security policies and should not bypass security controls. Too many steps in a "render useless" procedure will contribute to a potential compromise of valuable information. All of these determinants should be considered in the data security and human factors analysis in the development of the CDS/ES. 4 ASSESSMENT OF TECHNICAL CHALLENGES We now identify some of the key technical challenges that must be addressed in deigning and deploying the next generation of cross-domain solutions. Real-time processing. Delays in the sanitization and approval processes today impact mission effectiveness. Thus classical batch-oriented processing modes where information release request are queued up and sent in batches are severely limiting. What is needed are technologies and processes for the real-time processing of information release requests so as to increase their efficiency and timeliness. This involves eliminating multiple processing steps in the workflow where potential bottlenecks can appear. Automated risk-aware processing. Related to the above, is the need to increase the degree of automation in information analysis and release. Ideally, the system should be tuned so that time-consuming (but highly assured) human review etc, can be adjusted based on prespecified tradeoffs between efficiency and the risk of accidental leakage (from false negatives). So what is needed is a risk-aware information processing model which based on content sensitivity, environmental threat posture and recipient risk, will balance the degree of automation vs. human review so as to increase efficiency of workflows. Support for dynamic and efficient policy updates. We need to increase the flexibility, speed and efficiency with which policy changes can be incorporated and certified within CD devices. This requires (1) the ability to rapidly update policy rules without necessarily going back to the device manufacturer or vendor and (2) to get such rules validated and certified without laborious manual review. Higher-level and more interoperable policy specification and analysis. Policy specification is still relatively low-level, cumbersome and requires knowledge of vendor-specific filter specification languages etc. So there is a crucial need to raise the level of abstraction for policy specification so as to enable vendor-neutral and interoperable specification of cross-domain policies. This will also be a catalyst for the creation of automated policy analysis and validation tools and for more efficient certification approaches. Initial approaches to standardizing and making policy specification have been pursued from a bottom up [BRAY] as well as top-down [CDL 09] direction. However, further work needs to be done to integrate such approaches. Need to support complex (compound) data types. The goal here is to support complex data types such as compound documents so that the process of sanitization does not break the original data model. Traditionally many cross-domain devices such as guards and filters have assumed a simple flat data model consisting of simple messages. This has to evolve so that XML schemas and other tagged data representations and metadata can be supported. Individual elements in a complex structure need to be sanitized and regraded and access controlled at the recipient end. Need to improve traceability and accountability. The ability to provide post-release traceability and accountability to the sender of previously released information will go a long way in providing the flexibility and control some organizations desire, but cannot obtain from current technology. Also important is for the sender to have the ability to control the redissemination of the released information as well as revoke previously released information. Distributed and federated processing with miniaturized devices. Cross domain devices today are largely centralized devices placed at central egress and ingress points in the network topology of individual organizations. On the tactical edge, a more distributed and lightweight cross-domain device architecture may be more appropriate so as to provide the necessary flexibility and agility. Individual devices carried by the warfighter may have to Page 5 of7 Authorized licensed use limited to: Defence Science and Technology Group (DSTG). Downloaded on January 25,2023 at 00:00:24 UTC from IEEE Xplore. Restrictions apply. analyze and sanitize information. This brings to the forefront the challenges related to management and coordination of such devices. Remote management. Inherent to any distributed architecture is the need for remote management. Thus the technical challenge here is to design the remote management protocols and the necessary management plane that can be rapidly deployed over a mixture of wired and wireless networks in the battlefield. Remote management may include routine administrative functions such as security administration and patch updates, but also include the push of policy updates to devices and semiautomated recertification of such devices. Multi-party and third-party release and dissemination scenarios. The challenge here is to go beyond point-topoint cross-domain information release scenarios and support situations where there are varying and asymmetrical trust relationships between the various parties. This will be common place as coalition partners increasingly share information in real time. For example, A may release to B but B may not release that information to C without further sanitizations etc. A traditional cross domain solution providing services is depicted in Figure 4-1. Single Domain . , -, We also need to explore the use of the Service Oriented Architecture (SOA) framework. The use of SOA for policy distribution, providing centralized DoD policy management could be leveraged. This approach has the advantages of avoiding some of the issues of proper protection of data flow policy when the system is off or being transported. Methods of caching and distribution could also be investigated. There are many applications today where a warfighter and associated equipment at an unclassified level have data that need to be processed by classified algorithms. SOA or CORBA type interfaces could be used to provide the ability to upgrade the data, have it processed, and then down graded for use in the unclassified environment. This would provide valuable classified processing without exposing the classified algorithms to the unclassified environment. Standard architecture and interfaces for CDSIES would facilitate CDS/ES development and interaction. The SELinux initiative has proven to be a good starting point for providing a foundational architecture through a set of Linux Security Modules that enforce MAC and security policy enforcement that can be readily leveraged to develop CDSIES. More work still needs to be done to set standards for CDSIES interfaces. Interface standards including mutual authentication and service discovery need to be established. A standard policy language with a method of rich constraint expression should be established. This would provide the benefits of policy development to be conducted independent of CDSIES system development and would permit a more DoD centric management of CDS/ES policy [PCIM, CDL 08]. Figure 4-1 Existing CDS deployments 5 tactical networks, we cannot effectively fight. As our information systems provide increased reach back to intelligence and information, we in tum extend and expand the attack surface to the battlefield as well. As communications architectures evolve to support the warfighter, the number of avenues of refinement for CDS/ES evolution will naturally precipitate out. CONCLUSIONS Information systems used to process information and automate tasks have evolved from standalone computers the size of a building to networked, mobile, even to wearable devices at the tactical edge. In order to ensure the confidentiality, integrity, availability, and nonreputability of this information for the warfighter at the tactical edge, CDSIES must evolve as well. The need for information assurance is no more important than on the battlefield. As reliance upon information increases, so does need for automation. Without secure and available Continued work on standardizing data flows including the integration of newer DoD message formats would benefit producers and consumers of information that need to be transferred between domains. Joint Variable Message Format (JVMF) has been used in the past, and would be a good starting point. To support a common understanding of military command and control data the Multilateral Interoperability Program (MIP) has defined a data model to allow consistent sharing of command control and situational awareness data. This data model is called the Page 60f7 Authorized licensed use limited to: Defence Science and Technology Group (DSTG). Downloaded on January 25,2023 at 00:00:24 UTC from IEEE Xplore. Restrictions apply. Joint Consultation, Command and Control Information Exchange Data Model (JC3IEDM). The development of a XML based encapsulation of NMF and JC3IEDM has introduced significant improvement in facilitating and standardizing message transfer between domains [JC3IEDM]. CDS/ES standards for more real-time determinism (as much as can be afforded given CDSIES constraints) across the CDSIES. Standard network features as Resource Reservation Protocol (RSVP) or Quality of Service extensions in the policy and data languages as well as CDS/ES architecture should be explored . Lastly, current DoD infrastructure is plagued with single point failures, inefficiencies, and a lack of scalability methods for CDSIES. Methods to combat these issues should be researched. CDSIES brokerage , a CDSIES specific extension to routing protocols, and other network approaches should be examined. Recent advances in virtualization technology could also provide some relief to CDSIES in the areas of deployment, scalability, SWaP, mobility, and reliability . The emerging direction for tactical CDS is shown in Figure 5-1. ..,. " __ Multiple _Domains ~_ ,. .<-~ advances in the ability to deal with complex data and lastly improved user interfaces to enable the war fighter to make timely, safe and secure decisions. REFERENCES [BRAY] Boyd Fletcher, Bray: A Configuration and Validation Tool for DFCFs, CDS Technical Forum June 2009 [CJCSI] CJCSI 6211.02C Defense Information System Network (DISN): Policy and Responsibilities [CDL 09] Roshan K. Thomas, Giovanni Russello , Simon Tsang, Realizing the CDL Cross-Domain Language in the Ponder2 Policy Framework: Experiences and Research Directions, IEEE Policies for Distributed Systems and Networks 2009, July 2009 [CDL 08] Roshan K. Thomas and Simon Tsang, CDL: A Language for Specifying High-level Cross-Domain Security Policies, MILCOM, 2008 [IAMANET] D. Scott Alexander, Brian DeCleene , Jason Rogers, Pete Sholander, Requirements and architectures for Intrinsically Assurable Mobile Ad hoc Networks MILCOM 2008, November 2008 y~ [JC3IEDM] Joint Consultation, Command and Control Information Exchange Data Model 3.1, December 2006 [On Point] Gregory Fontenot, E.J. Degen, David Tohn, Operation IRAQI FREEDOM Study Group, The United States Army in Operation IRAQI FREEDOM, 2004 [MITRE] Elaine Caddick, Nancy Reed, Cross Domain Solutions : An Overview of Technology, Policy and Process , October 2005 [Owl] Owl Computing Cross-Domain Forum, Dr. Ron Mraz, Owl President & CTO, April 15, 2009 Man agemenUReview [PC1M] Policy Core Engineering Task Figure 5-1 Emerging CDS/ES solutions of the future Information Force, Model, RFC Internet 3060, http://www.ietf.org/rfc/rfc3060.txt This paper will hopefully serve as a starting point to explore new directions for the next generation of CDSIES solutions. These new directions will have to tackle issues ranging from higher-level and more interoperable policy specifications, more efficient policy updates and management, real-time processing, scalability and faulttolerance improvements through distributed processing, miniaturization of cross-domain platforms and devices, [CD Inventory] Unified Cross Domain Management Office, Cross Domain Inventory ver 3.0, May 2009 [Sim] R. Simmons, XML Schema Guidance for Cross Domain Security Policy Enforcement, MP 04W0000204, September 2004 Page 7 of7 Authorized licensed use limited to: Defence Science and Technology Group (DSTG). Downloaded on January 25,2023 at 00:00:24 UTC from IEEE Xplore. Restrictions apply.