Operational Feasibility of Adversarial Attacks Against Artificial Intelligence

by Li Ang Zhang, Gavin S. Hartnett, Jair Aguirre, Andrew J. Lohn, Inez Khan, Marissa Herron, Caolionn O'Connell

A large body of academic literature describes myriad attack vectors and suggests that most of the U.S. Department of Defense's (DoD's) artificial intelligence (AI) systems are in constant peril. However, RAND researchers investigated adversarial attacks designed to hide objects (causing algorithmic false negatives) and found that many attacks are operationally infeasible to design and deploy because of high knowledge requirements and impractical attack vectors. As the researchers discuss in this report, there are tried-and-true nonadversarial techniques that can be less expensive, more practical, and often more effective. Thus, adversarial attacks against AI pose less risk to DoD applications than academic research currently implies. Nevertheless, well-designed AI systems, as well as mitigation strategies, can further weaken the risks of such attacks.

更多

Key Findings

Adversarial attacks designed to hide objects from AI pose less risk to DoD applications than academic research currently implies.

In the real world, such adversarial attacks are difficult to design and deploy because of high knowledge requirements and infeasible attack vectors; there are often less expensive, more practical, and more effective nonadversarial techniques available.

Fusing data and predictions across sensor modalities, signal-sampling rates, and image resolution can further mitigate the risk of adversarial attacks against AI.

Recommendations

DoD should assess how at-risk its AI models are to adversarial attacks by considering how adversaries can feasibly influence models. It should also assess how knowledge leaks about models can affect attack efficacy and estimate the costs associated with adversary actions.

DoD should maintain situational awareness of academic state-of-the-art techniques to attack AI in real-world scenarios and understand how these technologies feasibly affect concepts of operation for both itself and its adversaries.

DoD should develop robust AI models, preprocessing techniques, and proper data fusion systems to vastly increase the resource costs to an adversary to perform an attack.

DoD should invest in responsive support teams for AI systems to quickly detect, identify, and respond to adversarial threats.

收起

Deception Detection

by Marek N. Posard, Christian Johnson, Julia L. Melin, Emily Ellinger, Hilary Reininger

A group of RAND Corporation researchers found that machine-learning (ML) models can identify signs of deception during national security background check interviews. The most accurate approach for detecting deception is an ML model that counts the number of times that interviewees use common words.

更多

How the Experiment Worked

The 103 participants read a story about Edward Snowden leaking classified information from the National Security Agency in 2013.

Participants were randomly assigned to read the same story, but the story was presented as either a news report or a memo with markings indicating it contained sensitive information.

Participants were assigned to one of two groups for interviews: One group was told to lie about what they had read, and the other was told to tell the truth.

Former law enforcement officers interviewed participants via video teleconference (VTC) and text-based chat in randomized order.

RAND researchers used the interview and chat transcripts to train several ML models to see whether the models could distinguish the liars from the truth-tellers.

收起

RAND — Making a Difference

For 75 years, decisionmakers around the world have turned to RAND for help solving the most complex and consequential public policy challenges. Through research and analysis that is empirical, nonpartisan, and objective, we strive to make communities throughout the world safer and more secure, healthier and more prosperous.

更多

The mix of studies, singular accomplishments, and streams of research and analysis included in this book exemplifies RAND's range, originality, and influence. The selections are listed in no particular order but are intended to reflect the breadth of our increasingly diversified research agenda. Taken together, the selections reveal a common motif: our ability to have a positive impact on the world by applying rigorous and objective analysis to challenging problems. That aspiration has guided RAND ever since it was established.

If you take one thing away from this book, perhaps it will be this: Public policy need not be inaccessible. It is about real people, real places, real organizations, real solutions.

收起

ELIZABETH M. BARTELS, ADAM R. GRISSOM, RUSSELL HANSON, CHRISTOPHER A. MOUTON

OCEANS 17 Tabletop

Exercise

更多

Findings and Recommendations

收起

Driving to Safety

How Many Miles of Driving Would It Take to Demonstrate Autonomous Vehicle Reliability?

Nidhi Kalra, Susan M. Paddock

A Review of Selected International Aircraft Spares Pooling Programs

Lessons Learned for F-35 Spares Pooling

by Mark A. Lorell, James Pita

更多

The F-35 Joint Strike Fighter is a fifth-generation stealth fighter whose system development was shared by nine countries: the United States (for which the F-35 is being jointly procured by the U.S. Air Force, U.S. Navy, and U.S. Marine Corps), Australia, Canada, Denmark, Italy, the Netherlands, Norway, Turkey, and the United Kingdom. In 2012, these countries agreed that F-35 sustainment assets would be managed as a single global pool, which would be centrally managed by the F-35 production support manager. In the formal agreement establishing this pool, language was also included to allow participants to opt out of the global pool, should a nation desire to establish (and be willing to pay for) a separate stock of assets. This report reviews other selected historical and current international military aircraft spares pooling programs involving the United States, major European North Atlantic Treaty Organization allies, and other U.S. allies to identify lessons learned. The most successful historical programs were characterized by a single dominant partner that could establish the ground rules and resource allocation priorities, as well as control configuration. At the same time, successful programs made major efforts to ensure fairness, equality, and transparency for all partners, based on relative contribution and need.

Key Findings

Lessons Learned from Historical International Spares Pooling Initiatives

The negotiation of international spares pooling programs for common major weapon systems has been attempted many times since the 1960s, but they have proven to be very difficult to implement.

Major barriers include (1) security of supply and prioritization of scarce spares assets and (2) configuration management and management of innovators versus laggards. These two challenges can be difficult to overcome unless one partner plays a dominant role in the pooling arrangements.

Conflicting industrial, technological, and economic interests and objectives were a major challenge, especially for all-European programs or programs in which all the partners were roughly of equivalent size and influence.

The most successful programs were characterized by a single dominant partner that could establish the ground rules and resource allocation priorities, as well as control aircraft configuration. At the same time, successful programs made major efforts to ensure fairness, equality, and transparency for all partners, based on relative contribution and need.

Policy Measures That Facilitate Spares Pooling

Successful programs have mitigated the problem of U.S. export control laws and regulations by retaining U.S. ownership of all spares except when they are installed in ally-owned aircraft.

All aircraft should conform to U.S. configuration standards, with the proviso that the U.S. Air Force pays for the nonrecurring costs of upgrading the aircraft.

Incentives for the prime contractor to meet international fleetwide performance metrics and priorities require a splitting out of the smaller foreign fleets with separately calculated metrics so that these fleets are serviced with the same priority as the larger U.S. fleet.

Table of Contents

Chapter One

Introduction and Overview

Chapter Two

NATO and European Experience with International Spares Pooling

Chapter Three

The C-17 Global Integrated Sustainment Partnership

Chapter Four

Conclusions

收起

Enabling Early Sustainment Decisions

Application to F-35 Depot-Level Maintenance

by John G. Drew, Ronald G. McGarvey, Peter Buryk

更多

The U.S. Air Force has long struggled to incorporate new weapon system logistics requirements and support system design considerations into its broader sustainment enterprise early in the acquisition process. To help inform Air Force decisionmaking with regard to sustainment sourcing, RAND Project AIR FORCE researchers explored and adapted lessons from the transaction cost accounting literature. The result is a powerful economic-based framework that has three primary benefits when it comes to addressing sustainment planning challenges: It is a repeatable, analytically driven decision tool that does not require large amounts of data; it considers repair source decisionmaking in the context of the broader Air Force enterprise; and it is potentially applicable to other aspects of sustainment planning, such as managing government-mandated repair sourcing mixes and informing other Air Force sustainment community responsibilities. This report demonstrates how the framework can be used to select among depot maintenance strategies by applying it to the F-35 Joint Strike Fighter, the largest acquisition program in U.S. Department of Defense history. Although the U.S. government will retain the capability to perform the range of depot-level repairs for the F-35, 40 percent of the workload — known as "above core" — can be considered for sourcing to an organic Air Force facility, another military service's facility, a foreign partner, or the private sector. The framework helps planners visualize program data and compare new acquisition programs with legacy Air Force systems. In this way, it offers the Air Force additional leverage in responding to technology developments and vetting contractors' engineering, reliability, and maintainability projections for new weapon systems.

Key Findings

Decisions About U.S. Air Force Weapon System Sustainment Often Do Not Consider the Effects on the Broader Sustainment Enterprise

Decisions about whether to use organic or contractor support for the long-term sustainment of a weapon system are often made within narrow organizational boundaries and do not consider the full range of capabilities in the enterprise to identify a more cost-effective option.

A review of legacy program data also showed that this decisionmaking can be data-intensive and often occurs in isolation and late in a weapon system's acquisition process.

History suggests that new aircraft systems generally act similarly to legacy aircraft systems at the technology/subsystem level. For this reason, the Air Force would benefit most from common sustainment strategies at this level as it moves forward with new acquisition programs.

The Air Force Would Benefit from a Framework to Support These Types of Decisions

U.S. Department of Defense guidance states that above-core depot workloads (those that the government has the option to outsource) should be assigned on the basis of a best-value determination, but this guidance does not specify how to determine "best value."

A RAND-developed economic-based framework addresses this gap by helping planners determine the "best value" when assigning above-core sustainment workloads, but such an analytic tool would have longer-term benefits for the Air Force. For example, it can help planners compare and discover commonalities across subsystems, identify the effects of new or emerging technology on Air Force sustainment strategies, and contribute to the Air Force's mission to shape its sustainment enterprise.

Recommendations

The Air Force should develop an enterprise sustainment strategy at the product support group or technology subsystem level and ensure that the appropriate sustainment infrastructure is in place.

From there, it should adopt a set of criteria for support system design sourcing decisions that weapon system program offices can apply early in the acquisition process. Such criteria would help determine whether the enterprise has the capability to sustain the system and, if not, whether it would be beneficial to develop the capability.

The RAND-developed framework offers the Air Force a helpful approach to start thinking more strategically about sustainment posture planning and illuminates the advantages of an enterprise-wide view of sustainment management. The Air Force should adopt the framework as it makes decisions regarding new acquisition programs, such as the F-35 Joint Strike Fighter, and as it addresses new technology developments affecting existing programs.

Table of Contents

Chapter One

Introduction

Chapter Two

TCA and Current Air Force Sustainment Efforts

Chapter Three

Application of the Framework to the F-35

Chapter Four

Conclusions and Potential Extensions of This Analysis

收起

Dire Strait?

Military Aspects of the China-Taiwan Confrontation and Options for U.S. Policy

by David A. Shlapak, David T. Orletsky, Barry Wilson

更多

Among the hottest flashpoints in the world today, the Taiwan Strait cannot be ignored by U.S. policymakers and diplomats. China regularly rattles its saber to intimidate Taiwan and influence U.S. policy but has thus far stopped short of overt military action. This report analyzes the steps Taiwan should take to bolster the odds in its favor should a conflict with the mainland occur and describes how the United States can most effectivelycontribute in both peace and crisis. The authors conclude that the United States and Taiwan can take a number of fairly simple and relatively inexpensive measures — including hardening air bases and other facilities and upgrading the air defense command and control system — that would significantly enhance Taiwan's ability to defend itself against a large-scale Chinese attack.

Table of Contents

Preface

Figures

Tables

Summary

Acknowledgements

Acknowledgments

Abbreviations

Chapter One

Introduction

Chapter Two

Scenario and Approach

Chapter Three

Issues and Implications

Chapter Four

Recommendations and Concluding Remarks

Appendix A

Some Thoughts on the PRC Missile Threat

Appendix B

Overview of the JICM

Bibliography

References

收起