Operator Suspicion and Human-Machine Team Performance Under Mission Scenarios of Unmanned Ground Vehicle Operation

Abstract

Emergent cyber-attack threats against cyber-physical systems can create potentially catastrophic impacts. The operators must intervene at the right moment when suspected attacks occur, without over-reliance on systems to detect the cyber-attacks. However, military operators are normally trained to trust, rather than suspect systems. We applied suspicion theory to explore how operators detect and respond to cyber-attacks against an unmanned ground vehicle (UGV) system in the operational context of a human-machine team (HMT). We investigated the relationships between the operator suspicion and HMT performance by conducting human-in-the-loop experiments on eight mission scenarios with 32 air-force officers. The experiment yielded a significant, negative relationship between operator suspicion and HMT performance (quantified both in terms of the desirability of decision response and the time to respond). Notably, operator suspicion increased with the combined effects of cyber-attacks and a sentinel alert but not with the alert alone. This finding was particularly meaningful for ''false-negative'' scenarios, in which no sentinel alert was sent despite cyber-attacks having occurred. Although the operators did not receive an alert, the operators grew more suspicious, seeking more information; it took longer for the operators to respond, and their decision responses were highly divergent (17.2% came with less-desirable responses, and 21.9% were considered instances of over-reliance). In contrast, in ''false-positive'' scenarios, 95.3% of the operator responses were highly desirable. This experiment has implications for the role of a sentinel alert in engineering trustworthy HMT systems so that the operators can quickly transition through state-suspicion to the most desirable decision. INDEX TERMS Suspicion, trust, human machine team, cyber security, human-in-the-loop simulation, unmanned ground vehicle control.

A Deep Learning-Enhanced Multi-Modal Sensing Platform for Robust Human Object Detection and Tracking in Challenging Environments

“…Autonomous vehicles, including unmanned aerial vehicles (UAVs) [1][2][3] and unmanned ground vehicles (UGVs) [4], have found extensive applications in agriculture [5], data acquisition [6], and search and rescue due to their mobility and operational simplicity. One significant capability desired in these search and surveillance scenarios is the ability of autonomous vehicles to recognize human subjects' actions and respond accordingly.…”

Assessing Current HMI Designs and Exploring AI Potential for Future Air-Defence System Development

“…A challenge, however, is a lack of theoretical understanding of how humans interact with machines in work contexts e.g. in terms of risk management and resiliency [11] [12] [13]. In the military, decisions have to be made within tight time schedules, often with incomplete information.…”

Mechanism Analysis of Smart Cue on Aircraft for Loss of Control Mitigation

“…Reference [19] proposed a stable adaptive approach which is used in the simultaneous sensor, actuator, and slow-engine gain failure accommodation of the flight control. In the research of the pilot-aircraft interface [20], [21], Klyde et al [22] proposed a smart adaptive flight effective cue (SAFE-Cue), namely, a smart inceptor implemented in the aircraft. The SAFE-Cue system features an adaptive command path gain to mitigate oscillation tendencies and an inceptor force-feedback cue to alert the pilot that the system is active.…”