Unmanned Aerial Systems (UAVs, Drones), initially known only for their military applications, are getting increasingly popular in the civil sector as well. Over the military canvas, drones have already proven themselves as a potent force multiplier through unmanned, round-the-clock, long-range and high-endurance missions for surveillance, reconnaissance, search and rescue, and even armed combat applications. With the emergence of the Internet of Things (IoT), commercial deployments of drones are also growing exponentially, ranging from cargo and taxi services to agriculture, disaster relief, risk assessment and monitoring of critical infrastructures. Irrespective of the deployment sector, drones are often entrusted to conduct safety, time and liability critical tasks, thus requiring secure, robust and trustworthy operations. In contrast, the rise in UAVs’ demand, coupled with market pressure to reduce size, weight, power and cost (SwaP-C) parameters, has caused vendors to often ignore security aspects, thus inducing serious safety and security threats. As UAVs rely on Global Positioning System (GPS) for positioning and navigation, they can fall prey to GPS jamming and spoofing attacks. The vulnerability of GPS to spoofing has serious implications for UAVs, as victim drones using civil GPS can be misdirected or even completely hijacked for malicious intents, as already demonstrated in several academic research efforts using commercially available GPS spoofing hardware. Beside UAVs, GPS spoofing attacks are equally applicable to other GPS-dependent platforms, including manned aircraft, ground vehicles, and cellular systems. This paper conducts a comprehensive review of GPS spoofing threats, with a special focus on their applicability over UAVs and other GPS-dependent mobile platforms. It presents a novel taxonomy of GPS spoofing attacks and critically analyzes different spoofing techniques based upon placement of spoofing device, attack stealthiness, attack methodologies, and objectives of the attacker. We also discuss some of the recent experiments from open literature which utilized commercially available hardware for successfully conducting spoofing attacks.

Visual Language Navigation (VLN) is a rapidly evolving, cross-disciplinary field that has witnessed a phenomenal rise since its inception in recent past. Leveraging the power of deep learning, it aims to endow auditory and visual perceptions to embodied agents, thereby enabling them to interpret the multi-modal cues. Although the end objective remains solitary, the various ways this problem is approached in the larger context of artificial intelligence differs substantially in scope. Consequently, various contributions, ranging from architectural innovations and novel models, improved training strategies, datasets, methodologies, etc. have been proposed in the literature. Therefore, this work aims to provide a comprehensive review of scientific contributions and advancements by categorizing the literature into four primary abstraction layers: frameworks, models, auxiliaries, and tasks. Within each abstraction layer, the literature is further sub-categorized and then analyzed for notable trends, strengths, weaknesses and limitations as well as cross-compared for respective advantages and limitations. The simulation environments and datasets have been consolidated and unique insights are presented into each along with a summary of their advantages and disadvantages. Through extensive analysis, crucial challenges that remain in VLN are identified, guiding future efforts toward more intelligent and adaptable navigation agents.