Implementation of Tensor Neural Networks

Tensor networks are a quantum inspired technique that provide useful mechanisms for solving hard problems, usually in the realm of quantum mechanics and many-body systems, through the manipulation and operation of tensors. The main challenge underlying this project is the productionalization of a quantum-inspired deep learning algorithm which currently exists as a proof of concept (POC). That POC was specifically built for a single user’s needs, applied to the fair pricing of financial derivatives.

Assessing the Effect of Hazard Perception Training on Driving Ability

Improving driving safety is an ongoing challenge with far reaching consequences for Canadians. One method of improving driving safety is by improving driver performance with hazard perception training. Hazard perception is a driver's ability to anticipate dangers in the road environment and is predictive of the likelihood of being in a crash. Several countries including the United Kingdom, Netherlands, and Australia have introduced this training as part of getting a license but there is no current equivalent in Canada.

A framework to enhance deep learning systems’ trustworthiness against Out of Distribution examples

In the past decade, deep learning models have demonstrated their highest performance for a variety of tasks. These models outperformed classical machine learning models and even humans in terms of performance and accuracy. However, previous research indicated that these models are vulnerable to out-of-distribution and adversarial inputs. Ideally, these inputs should be rejected by the deep learning model, but the deep learning model generates confident outcomes for it.

Tiresias: Client Private Malware Protection

Tiresias is a client private solution to malware protection and threat intelligence. Tiresias allows a user to put all their incoming files in a cryptographically secure Data Chest locally. After sending the Data Chest to our cloud environment, our AI scans and infers if it is malicious without seeing the actual file content from the Data Chest. This method protects the client data privacy and confidentiality. The Data Chest is a novel research outcome at the Queen’s School of Computing.

Bioactive nanoparticles-based domiciliary oral gel with antimicrobial, immunomodulatory and antioxidant activity for periodontal therapy

Periodontal diseases are among the most common diseases in the population. According to the Canadian Dental Association, around 7 in 10 Canadians suffer from some form of gum disease at some point in their lives. They are characterized by inflammation and destruction of tooth supporting structures. Their progression is determined by the interaction between microbes and the host immune response. Recently, locally administered therapies have been proposed as a support to the gold standard treatment scaling and root planning, but no consensus on their effectiveness is present.

A novel phishing detection approach using Fuzzy Logic and Deep Learning

People are switching from traditional shopping to internet commerce as Internet access increases quickly. So nowadays people are becoming more dependent on e-commerce-based websites. On the other hand, instead of robbing businesses like banks and stores, modern thieves now use the anonymous internet architecture to track down their victims online. Hackers are employing new strategies, such as phishing, to deceive their victims by creating fake websites to collect sensitive data, such as account numbers, usernames, and passwords.

Anonymous Age Verification Using Electrocardiogram (ECG) Obtained from Smart Wearables

Age-verification mandatory procedure for delivering certain services and products. Traditionally, identification documents have been a common mechanism of age-verification. However, this current strategy is subject to certain risks regarding privacy protection and online forgery. This demonstrates the value in anonymous age verification schemes using biometrics. Considering its age-dependent attributes, Electrocardiogram (ECG) is a potential solution. Preliminary experiments have been conducted regarding age estimation- classification from ECG obtained in clinical settings.

Preserving Privacy at Edge Devices

The aim of this project is to develop an application that can proactively protect users from identity theft and create awareness around safe digital practices. For this, we will be developing novel ways of extracting utility out of data and helping users to maintain a least-risk profile score. Most of the computations will be done on edge devices and computations will be prioritized based on value extractions. This research will help to bridge the gap between the accuracy and efficiency of the models to preserve privacy.

MEV protection through delayed execution and time-locked puzzles

Blockchain technology is, by nature, transparent and decentralized. However, transparency can sometimes be a direct threat to the system's decentralization when most of the profitable transactions get stolen by several high-power entities. This de-incentivizes new users from using the system and renders the whole system useless. In this project, we aim to guard decentralization against destructive and unfair effects of transparency. We reason that transparency can harm the system when it enables users to easily front-run each other.

Evaluation of a green graphite based LTO battery

With millions of lithium batteries in the marked and billions yet to be, the question is can we produce a battery that can last 100 years, cannot catch fire, work in freezing conditions, charge 10 times faster and yet made from ecofriendly materials like graphite, rubber, wallpaper paste and paper? The answer is yes. Our goal is to attempt to build this novel eco-friendly battery and benchmark it with its equivalent traditionally produced counterpart. We anticipate the ecofriendly battery to perform on par but at a much lower cost of production.