skip to main content


The NSF Public Access Repository (NSF-PAR) system and access will be unavailable from 5:00 PM ET until 11:00 PM ET on Friday, June 21 due to maintenance. We apologize for the inconvenience.

Title: Targeted Attack on Deep RL-based Autonomous Driving with Learned Visual Patterns
Recent studies demonstrated the vulnerability of control policies learned through deep reinforcement learning against adversarial attacks, raising concerns about the application of such models to risk-sensitive tasks such as autonomous driving. Threat models for these demonstrations are limited to (1) targeted attacks through real-time manipulation of the agent's observation, and (2) untargeted attacks through manipulation of the physical environment. The former assumes full access to the agent's states/observations at all times, while the latter has no control over attack outcomes. This paper investigates the feasibility of targeted attacks through visually learned patterns placed on physical objects in the environment, a threat model that combines the practicality and effectiveness of the existing ones. Through analysis, we demonstrate that a pre-trained policy can be hijacked within a time window, e.g., performing an unintended self-parking, when an adversarial object is present. To enable the attack, we adopt an assumption that the dynamics of both the environment and the agent can be learned by the attacker. Lastly, we empirically show the effectiveness of the proposed attack on different driving scenarios, perform a location robustness test, and study the tradeoff between the attack strength and its effectiveness Code is available at Targeted-Physical-Adversarial-Attacks-on-AD  more » « less
Award ID(s):
2038666 1925403
Author(s) / Creator(s):
; ; ;
Date Published:
Journal Name:
2022 International Conference on Robotics and Automation (ICRA)
Page Range / eLocation ID:
10571 to 10577
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. In face of an increasing number of automotive cyber-physical threat scenarios, the issue of adversarial destabilization of the lateral motion of target vehicles through direct attacks on their steering systems has been extensively studied. A more subtle question is whether a cyberattacker can destabilize the target vehicle lateral motion through improper engagement of the vehicle brakes and/or anti-lock braking systems (ABS). Motivated by such a question, this paper investigates the impact of cyber-physical attacks that exploit the braking/ABS systems to adversely affect the lateral motion stability of the targeted vehicles. Using a hybrid physical/dynamic tire-road friction model, it is shown that if a braking system/ABS attacker manages to continuously vary the longitudinal slips of the wheels, they can violate the necessary conditions for asymptotic stability of the underlying linear time-varying (LTV) dynamics of the lateral motion. Furthermore, the minimal perturbations of the wheel longitudinal slips that result in lateral motion instability under fixed slip values are derived. Finally, a real-time algorithm for monitoring the lateral motion dynamics of vehicles against braking/ABS cyber-physical attacks is devised. This algorithm, which can be efficiently computed using the modest computational resources of automotive embedded processors, can be utilized along with other intrusion detection techniques to infer whether a vehicle braking system/ABS is experiencing a cyber-physical attack. Numerical simulations in the presence of realistic CAN bus delays, destabilizing slip value perturbations obtained from solving quadratic programs on an embedded ARM Cortex-M3 emulator, and side-wind gusts demonstrate the effectiveness of the proposed methodology. 
    more » « less
  2. Malicious software (malware) is a major cyber threat that has to be tackled with Machine Learning (ML) techniques because millions of new malware examples are injected into cyberspace on a daily basis. However, ML is vulnerable to attacks known as adversarial examples. In this article, we survey and systematize the field of Adversarial Malware Detection (AMD) through the lens of a unified conceptual framework of assumptions, attacks, defenses, and security properties. This not only leads us to map attacks and defenses to partial order structures, but also allows us to clearly describe the attack-defense arms race in the AMD context. We draw a number of insights, including: knowing the defender’s feature set is critical to the success of transfer attacks; the effectiveness of practical evasion attacks largely depends on the attacker’s freedom in conducting manipulations in the problem space; knowing the attacker’s manipulation set is critical to the defender’s success; and the effectiveness of adversarial training depends on the defender’s capability in identifying the most powerful attack. We also discuss a number of future research directions. 
    more » « less
  3. The NTT (Nippon Telegraph and Telephone) Data Corporation report found that 80% of U.S. consumers are concerned about their smart home data security. The Internet of Things (IoT) technology brings many benefits to people's homes, and more people across the world are heavily dependent on the technology and its devices. However, many IoT devices are deployed without considering security, increasing the number of attack vectors available to attackers. Numerous Internet of Things devices lacking security features have been compromised by attackers, resulting in many security incidents. Attackers can infiltrate these smart home devices and control the home via turning off the lights, controlling the alarm systems, and unlocking the smart locks, to name a few. Attackers have also been able to access the smart home network, leading to data exfiltration. There are many threats that smart homes face, such as the Man-in-the-Middle (MIM) attacks, data and identity theft, and Denial of Service (DoS) attacks. The hardware vulnerabilities often targeted by attackers are SPI, UART, JTAG, USB, etc. Therefore, to enhance the security of the smart devices used in our daily lives, threat modeling should be implemented early on in developing any given system. This past Spring semester, Morgan State University launched a (senior) capstone project targeting undergraduate (electrical) engineering students who were thus allowed to research with the Cybersecurity Assurance and Policy (CAP) center for four months. The primary purpose of the capstone was to help students further develop both hardware and software skills while researching. For this project, the students mainly focused on the Arduino Mega Board. Some of the expected outcomes for this capstone project include: 1) understanding the physical board components, 2) learning how to attack the board through the STRIDE technique, 3) generating a Data Flow Diagram (DFD) of the system using the Microsoft threat modeling tool, 4) understanding the attack patterns, and 5) generating the threat based on the user's input. To prevent future threats and attacks from taking advantage of systems vulnerabilities, the practice of "threat modeling" is implemented. This method allows the analysis of potential attackers, including their goals and techniques, while also providing solutions and mitigation strategies. Although Threat modeling can be performed throughout the development of a system, implementing it during developmental stages will prevent further problems in the future. Threat Modeling is crucial because it will help identify any potential threat before it propagates in the system. Identifying threats and providing countermeasures will save both time and money while also keeping the consumers safe. As a result, students must grow to understand how essential detecting and preventing attacks are to protect consumer information systems and networks. At the end of this capstone project, students should take away hands-on skills in cyber defense. 
    more » « less
  4. The safety-critical nature of vehicle steering is one of the main motivations for exploring the space of possible cyber-physical attacks against the steering systems of modern vehicles. This paper investigates the adversarial capabilities for destabilizing the interaction dynamics between human drivers and vehicle haptic shared control (HSC) steering systems. In contrast to the conventional robotics literature, where the main objective is to render the human-automation interaction dynamics stable by ensuring passivity, this paper takes the exact opposite route. In particular, to investigate the damaging capabilities of a successful cyber-physical attack, this paper demonstrates that an attacker who targets the HSC steering system can destabilize the interaction dynamics between the human driver and the vehicle HSC steering system through synthesis of time-varying impedance profiles. Specifically, it is shown that the adversary can utilize a properly designed non-passive and time-varying adversarial impedance target dynamics, which are fed with a linear combination of the human driver and the steering column torques. Using these target dynamics, it is possible for the adversary to generate in realtime a reference angular command for the driver input device and the directional control steering assembly of the vehicle. Furthermore, it is shown that the adversary can make the steering wheel and the vehicle steering column angular positions to follow the reference command generated by the time-varying impedance target dynamics using proper adaptive control strategies. Numerical simulations demonstrate the effectiveness of such time-varying impedance attacks, which result in a non-passive and inherently unstable interaction between the driver and the HSC steering system. 
    more » « less
  5. Automated Lane Centering (ALC) systems are convenient and widely deployed today, but also highly security and safety critical. In this work, we are the first to systematically study the security of state-of-the-art deep learning based ALC systems in their designed operational domains under physical-world adversarial attacks. We formulate the problem with a safetycritical attack goal, and a novel and domain-specific attack vector: dirty road patches. To systematically generate the attack, we adopt an optimization-based approach and overcome domain-specific design challenges such as camera frame interdependencies due to attack-influenced vehicle control, and the lack of objective function design for lane detection models. We evaluate our attack on a production ALC using 80 scenarios from real-world driving traces. The results show that our attack is highly effective with over 97.5% success rates and less than 0.903 sec average success time, which is substantially lower than the average driver reaction time. This attack is also found (1) robust to various real-world factors such as lighting conditions and view angles, (2) general to different model designs, and (3) stealthy from the driver’s view. To understand the safety impacts, we conduct experiments using software-in-the-loop simulation and attack trace injection in a real vehicle. The results show that our attack can cause a 100% collision rate in different scenarios, including when tested with common safety features such as automatic emergency braking. We also evaluate and discuss defenses. 
    more » « less