DR-NTU (Digital Repository of NTU)
Not a member yet
116018 research outputs found
Sort by
SoK: a taxonomic analysis of DeFi rug pulls: types, dataset, and tool assessment
No full textRug pulls present a critical threat in Decentralized Finance (DeFi), causing substantial financial losses and eroding ecosystem trust. Despite research advances, effective detection remains hampered by fragmented taxonomies, limited datasets, and inadequate tool evaluations. Through systematic analysis of academic and industry sources, we develop a comprehensive taxonomy of 35 distinct rug pull types, including 9 previously undocumented variants. Our analysis reveals significant detection gaps: existing datasets cover only 20% of known types, leading us to create an enhanced dataset of 2,391 instances that increases coverage to 82.9%. Evaluation of 13 detection tools shows substantial capability variation (25.7% to 62.9%), with 9 types completely undetectable. Most critically, tool performance degrades significantly when facing complex attacks, with maximum detection rates dropping from 55.6% for single-vector cases to 31.3% for compound scenarios. These findings provide essential insights for developing more robust security testing approaches for smart contract vulnerabilities in decentralized systems.Published versionThis work was supportedby NSFC Grant (No. W2412110)
Ultra-low voltage SRAM for energy minimisation in IoT devices
No full textIt is now more important than ever to optimise Static Random-Access Memory (SRAM) for ultra-low voltage operation due to the Internet of Things' (IoT) growing demand for energy-efficient computing. To reduce energy dissipation while preserving data stability and functional performance under low voltage constraints, this project explores several innovative bit cell topologies. Advanced configurations, such as PMOS-PMOS-NMOS (PPN) based designs and multi-threshold voltage (multi-Vt) schemes, were assessed alongside traditional SRAM structures like 6T, 8T, and 10T. The TSMC 65nm design kit was used to conduct circuit-level simulations to evaluate two important performance metrics: power dissipation and static noise margin (SNM). Because of its exceptional stability, the analysis determined that the 8T topology would make an appropriate foundational structure. This structure was then enhanced by combining an HRL-configured multi-threshold scheme with the PPN inverter topology, resulting in the proposed solution. The final design showed trade-offs, achieving a slight decrease in SNM in comparison to the base 8T configuration but an improvement in power efficiency when compared to standalone multi-Vt topologies. Additional simulations of read/write delay, access time, and leakage characteristics, as well as assistive techniques like word line (WL) boosting, could not be fully implemented due to time and tool constraints. However, WL boosting was examined theoretically and suggested as a promising avenue for further research, especially considering its capacity to resolve the long-standing trade-off between access speed and power efficiency. To summarise, the project lays a solid foundation for further study into ultra-low-voltage SRAM design, demonstrating the feasibility of integrating assistive technologies with bit cell advancements to meet the demanding energy and performance requirements of next-generation IoT systems.Bachelor's degre
Collisions of elliptic vortex rings upon flat walls
No full textHead-on collisions between elliptic vortex rings (EVRs) and walls were studied experimentally using planar laser-induced fluorescence visualisations and time-resolved particle image velocimetry. Aspect ratios of AR=2
and 4 EVRs at a Reynolds number of Re=4000
were used. Collision locations were based on four key axis-switching stages of freely translating EVRs, which would shed light upon how axis-switching behaviour and aspect ratio variations affect the collision outcomes. Results show that non-uniform circumferential induced velocities in both colliding EVRs produce different behaviours along major and minor planes, where vortex-stretching/compression and hence circumferential flows play key roles in the vortex dynamics. Non-uniform formations of secondary/tertiary EVRs also lead to varied entanglements around the primary EVRs. As such, secondary vortex rings form vortex loops that may congregate along the collision axis, depending on the exact collision location. Vortex-core trajectories show the net primary/secondary vortex-core movements result from a balance between EVR diameter expansion due to collision and EVR segment motions associated with the axis-switching stage at the point of collision. Confinement effects are also observed to dominate over aspect ratio effects when the collision occurs closest to the orifice. While increasing the aspect ratio leads to different vortex-stretching/compression behaviour and more varied vortex-core trajectories due to the greater non-uniform induced velocities, they could still be understood by the preceding interpretations. Finally, three-dimensional vortex flows are reconstructed based on the experimental results to further explain the flow mechanisms.Published versio
Investigating acoustic stereotypes of voice quality in protagonists and antagonists in film
No full textCharacters in film are anecdotally identified using acoustic stereotypes in terms of voice quality where antagonists are often associated with employing non-modal phonation such as creaky voice. However, little empirical research substantiates this informal observation, particularly in the domain of live-action English-medium films. This study seeks to quantitatively examine voice qualities of protagonist and antagonist characters in that domain using acoustic features such as Long-Term Average Spectrum (LTAS) and jitter, shimmer and Harmonics-to-Noise (HNR) to empirically determine if specific voice qualities are associated with character role. GAMMs was used to identify non-linear relationships between role and frequency for the spectra and k-means clustering and LMMs were used for the remaining three. Results showed that while LTAS showed statistical significance in antagonists and protagonists utilising different voice qualities, the cluster and LMMs did not corroborate this observation. This difference is attributed to the nature of the acoustic metrics, where only LTAS looks at the broad spectra of voice quality. Future studies can implement granular use of the acoustic metrics over the length of voice sample to corroborate results with LTAS.Bachelor's degre
Interactive learning on simple infrared transceiver
No full textIR communication is widely used in consumer electronics for wireless control. It can
be found in the devices we use daily, such as air-con and TV remotes. These systems
rely on a system called IR transceiver, which includes an IR emitter/transmitter LED
and IR receiver. The IR receiver detects modulated IR pulses send by the IR emitter
LED and converts them into electrical signals. This Final-Year Project report explores
the design and implementation of IR transceiver circuit in an IR Shooting Game, which consists of an IR Blaster and Target Board. Additionally, this project also aims to enhance learning through an interactive DIY hands-on kit named InnoKit in conjunction with a website that demonstrates the working principal of the IR transceiver system, and in addition guiding the learners on how to build the kit. The targeted audience can range from beginners to amateurs and from students to adults, arming them with the knowledge of how electronics in their essential daily appliances workBachelor's degre
Will AI enhance decision-making in the use of nuclear weapons?
No full textAI has been integrated into nuclear weapon doctrines to facilitate efficient autonomous decision-making. While speed is often crucial in military operations, decisions to launch weapons of mass destruction, such as nuclear weapons, require deliberate human intuition and intervention that surpasses calculated assessments generated by AI.Published versio
Buckling in constrained recovery in shape memory materials
No full textThis study explores constrained recovery in shape memory materials, highlighting how pre-strain
and pre-twist influence their behavior. Similar recovery patterns were observed across different
materials, despite varying stress magnitudes.Bachelor's degre
Design of wideband, high gain low noise OTA
No full textThe design of a nested-current-mirror (NCM) single-stage amplifier to advance its GBW-to power/area efficiency and CL drivability beyond the multi-stage designs, while preserving a rail-to-rail output swing. This design will be in the 0.18um CMOS technology, and should be able to work at 1.8V down to 1.2V power supply with 60 degrees phase margin, GBW of at least 1GHz, open-loop gain of about 80dB, with an input-referred noise of about 500nV/sq.rt.Hz at 1kHz. The driving capability of the amplifier is for 2pF, and the total current consumption of below <120uA if possible. The biasing is supplied through a constant Gm wide-swing cascode circuit such as to achieve a good PSRR for the entire design.Bachelor's degre
Adapting to change: the impact of COVID-19 on project management practice and the role of project managers
No full textThis research was inspired by the publication of “The COVID-19 Pandemic and Project Management Research” that called for the future researchers to explore and expand
Project management studies.
The focus of this Research is on the discovery of the potential changes incurred before and after Covid-19 in Project Management. As the world fell suddenly Covid spread
furiously without mercy across APAC in early 2020, NAFTA, Europe, Africa and the
Middle East. The supply chain for essential products and consumer goods almost came to a halt. Construction sectors which relied heavily on human resources were severely affected. Tourism and air travel suffered a sharp decline as borders across countries imposed strict restriction to travellers to decelerate and control the spread of Covid 19.
Major financial markets and stock exchange indices came crashing down, reacting to
news of fatality, restriction policies by government creating ill sentiment which brought about the "Bear Market". These events have brought the world economy to almost a standstill. However, the impact of Covid- 19 brought growth to the IT industry as
businesses rushed to bring about digitalization transformation. These sudden changes
provide opportunities and change to PM. From the traditional face-to-face (F2F) interactive way to a digital platform utilizing solutions and practices related to IoT. These changes require leaders to make alterations to their usual way of managing organization strategies that directly impact ongoing and new projects.
The research design uses literature review as its source. A selection of multiple research papers was obtained with search engines such as “Google Scholar” and “Google”. To ensure the quality of the selected research paper verification, this was conducted via “NTU Library” database to ascertain the papers were “peer reviewed”. To further enrich
and ascertain the content, supporting references and supplementary resource, reputable
consultation service company web site such as “McKinsey & Company” and officialMaster's degre
Evaluation of deformable image registration for contour propagation and dose accumulation in adaptive radiotherapy for prostate cancer
No full textAdaptive radiotherapy (ART) is a process whereby treatment plans are modified according to the latest patient anatomy, to ensure that the target coverage and dose sparing of organs at risk (OARs) are not compromised. Image registration, a technique to transform one image to another, plays an important role in ART. Due to the flexible nature of human tissues, deformable image registration (DIR) which offers the greatest degree of freedom has emerged as the most used method to transform patients’ images, including computed tomography (CT) and cone-beam CT scans. Some common applications of DIR in ART include synthetic CT generation, contour propagation and dose accumulation. For a reliable clinical translation of the DIR results, quality assurance (QA) must be performed to verify the registration accuracy. There are two types of metrics available for DIR QA: image-based metrics and deformation vector field (DVF)-based metrics. Image-based metrics comprise contour-based metrics and landmark technique.
While contour-based metrics have been widely studied and well established for DIR QA, this type of metrics require manual delineation of contours by radiation oncologists, rendering it time-consuming and subject to inter- or even intra-observer variability. In addition, several studies have suggested that contour-based metrics do not reflect the registration accuracy of regions within the contours, while landmark technique proves to be superior in this context. Meanwhile, DVF-based metrics offers the advantage of being readily available after registration, thus having the potential to expedite the DIR QA process. This thesis mainly addressed the DIR QA paradigm for prostate cancer patients in two aspects, contour propagation and dose accumulation. By quantifying and correlating the different types of metrics, we not only aimed to evaluate various DIR algorithms, but also to provide a linkage between these metrics, with the ultimate goal of speeding up the QA process and paving the way for a more robust QA system.
All the projects discussed in this thesis focused on three major organs in prostate RT, including prostate (target), bladder and rectum (OARs). The first part of the thesis dealt with the DIR application in contour propagation. A novel DVF-based metric, named normalised projection, was introduced. Along with normalised projection, inverse consistency error and Jacobian determinant were also studied. Pearson correlation tests were conducted between these three DVF-based metrics and two contour-based metrics, i.e., Dice similarity coefficient (DSC) and Hausdorff distance (HD). Subsequently, machine learning (ML) models were trained to predict DSC using DVF-based metrics. The models developed were linear regression (LR), Nu Support Vector Regression (NuSVR) and Random Forest Regressor (RFR). The second part of the thesis investigated the dose accumulation application. A deformable anthropomorphic pelvis phantom was fabricated via 3D printing, and implanted with markers. These markers provided the ground truth for measuring voxel mapping accuracy, through quantification of target registration error (TRE). Four DIR software – RayStation v10B, Velocity v4.1, Slicer, and Plastimatch – were analysed by computing the mean TRE and examining its correlation with DSC.
The normalised projection metric exhibited moderately positive and negative correlations with DSC and average HD, respectively, highlighting its predictive capability for contour-based metrics. The ability of ML models to extract important features further brought out the potential of DVF-based metrics in predicting DSC. Out of the three models trained, RFR performed the best, with the average mean absolute error lower than 0.07. As for dose accumulation, the in-house built phantom enabled simulation of realistic deformation scenarios and evaluation of voxel mapping accuracy. Comparison across four DIR software showed that RayStation had the highest voxel mapping accuracy. Its mean TRE ranged between 1.48 mm and 3.06 mm, which is close to or smaller than the AAPM recommended upper tolerance limit (3 mm). SlicerElastix (an extension in Slicer) displayed a remarkably strong negative correlation between mean TRE and DSC for all the organs considered.
In conclusion, this thesis demonstrated the feasibility of using DVF-based metrics to predict contour-based metrics, suggesting the potential for DVF-based metrics to streamline the DIR QA workflow for contour propagation. The design of deformable phantom with markers allowed the quantification and benchmarking of voxel mapping accuracy in terms of TRE, as well as an organ-wise correlation analysis between mean TRE and DSC. Future work could be done by making appropriate modifications to the deformable phantom, thereby enabling DIR validation for dose accumulation via dosimetric criteria at the prostate site.Doctor of Philosoph