A5053332315- Real-Time Systems Scheduling
- Embedded Systems Design Techniques
- Adversarial Robustness in Machine Learning
- Distributed systems and fault tolerance
- Advanced Neural Network Applications
- Petri Nets in System Modeling
- Anomaly Detection Techniques and Applications
- IoT and Edge/Fog Computing
- Domain Adaptation and Few-Shot Learning
- Advanced Malware Detection Techniques
- Indoor and Outdoor Localization Technologies
- Bluetooth and Wireless Communication Technologies
- Vehicle Dynamics and Control Systems
- Radiation Effects in Electronics
- IoT-based Smart Home Systems
- Real-time simulation and control systems
- Autonomous Vehicle Technology and Safety
- Age of Information Optimization
- Distributed and Parallel Computing Systems
- Hydraulic and Pneumatic Systems
- Intravenous Infusion Technology and Safety
- Tribology and Lubrication Engineering
- Network Security and Intrusion Detection
- Human Pose and Action Recognition
North Carolina State University
2023-2024
University of Central Florida
2021-2022
Over the last decade, machine learning (ML) and deep (DL) algorithms have significantly evolved been employed in diverse applications, such as computer vision, natural language processing, automated speech recognition, etc. Real-time safety-critical embedded Internet of Things (IoT) systems, autonomous driving UAVs, drones, security robots, etc., heavily rely on ML/DL-based technologies, accelerated with improvement hardware technologies. The cost a deadline (required time constraint) missed...
In Virtual Reality (VR), users typically interact with the virtual world using keyboard to insert keywords, surfing webpages, or typing passwords access online accounts. Hence, it becomes imperative understand security of keystrokes. this paper, we present VR-Spy, a keystrokes recognition method channel state information (CSI) WiFi signals. To best our knowledge, is first work that uses signals recognize in VR headsets. The key idea behind -Spy side-channel fine-granular hand movements...
Robot Operating System (ROS) is the most popular framework for developing robotics software. Typically, software safety-critical and employed in real-time systems requiring timing guarantees. Since first generation of ROS provides no guarantee, recent release its second generation, ROS2, necessary timely, has since received immense attention from practitioners researchers. Unfortunately, existing analysis ROS2 showed peculiar scheduling strategy executor, which severely affects response time...
Software reusability and system modularity are key features of modern autonomous systems. As a consequence, there is rapid shift towards hierarchical compositional architecture, as evidenced by AUTOSAR in automobiles ROS2 robotics. The resource-budget supply model widely applied the real-time analysis such Meanwhile, systems with multiple critical levels have received significant attention from research community industry. These designed execution budgets for system-critical levels. Existing...
Source-free domain adaptation (SFDA) is a popular unsupervised method where pre-trained model from source adapted to target without accessing any data. Despite rich results in this area, existing literature overlooks the security challenges of SFDA setting presence malicious owner. This work investigates effect adversary which may inject hidden behavior (Backdoor/Trojan) during training and potentially transfer it even after benign by victim (target owner). Our investigation current reveals...
The second generation of robot operating system (ROS 2) received significant attention from the real-time research community, mostly aiming at providing formal modeling and timing analysis. However, most current efforts are limited to default scheduling design schemes ROS 2. unique policies maintained by 2 significantly affect response time acceptance rate workload schedulability. It also invalidates adaptation rich existing results related nonpreemptive (and limited-preemptive) problems in...
Recent studies have revealed the vulnerability of deep neural networks (DNNs) to various backdoor attacks, where behavior DNNs can be compromised by utilizing certain types triggers or poisoning mechanisms. State-of-the-art (SOTA) defenses employ too-sophisticated mechanisms that require either a computationally expensive adversarial search module for reverse-engineering trigger distribution an over-sensitive hyper-parameter selection module. Moreover, they offer sub-par performance in...
Studies on backdoor attacks in recent years suggest that an adversary can compromise the integrity of a deep neural network (DNN) by manipulating small set training samples. Our analysis shows such manipulation make model converge to bad local minima, i.e., sharper minima as compared benign model. Intuitively, be purified re-optimizing smoother minima. However, na\"ive adoption any optimization targeting lead sub-optimal purification techniques hampering clean test accuracy. Hence,...
Studies on backdoor attacks in recent years suggest that an adversary can compromise the integrity of a deep neural network (DNN) by manipulating small set training samples. Our analysis shows such manipulation make model converge to bad local minima, i.e., sharper minima as compared benign model. Intuitively, be purified re-optimizing smoother minima. However, naïve adoption any optimization targeting lead sub-optimal purification techniques hampering clean test accuracy. Hence, effectively...
The success of a deep neural network (DNN) heavily relies on the details training scheme; e.g., data, architectures, hyper-parameters, etc. Recent backdoor attacks suggest that an adversary can take advantage such and compromise integrity DNN. Our studies show model is usually optimized to bad local minima, i.e. sharper minima as compared benign model. Intuitively, be purified by reoptimizing smoother through fine-tuning with few clean validation data. However, all DNN parameters often...
Real-time systems are characterized by strict timing constraints represented deadlines. Some tight, such that jobs finish their execution right at the deadlines in worst case, while others may not be so tight. Static slack is a concept captures non-tightness, and it can often "stolen" to handle additional aperiodic job requests, task suspensions, occasional overruns. This paper identifies an interesting direct correlation between worst-case response time (WCRT) static deadline-driven...