MSIT Theses and Dissertations (2019)

Permanent URI for this collection

http://hdl.handle.net/11071/6621

Browse

Now showing 1 - 20 of 25

An Algorithm for inferring consumer-to-consumer trust on twitter
(Strathmore University, 2019) Loyatum, Dennis
Trust amongst users engaged in consumer-to-consumer (C2C) e-commerce on Twitter as well as other social media platforms has been on the decline. The cost effective manner and timely delivery of C2C content makes it possible to reach a wider consumer base across the globe. However, this is under threat partly because of the risk of being scammed by other consumers on these platforms and the uncertainty related to this kind of e-commerce. Social media platforms such as Twitter are experiencing a decline in active user partly because of misuse of their platform. Twitter features can be used to build a consumer-to-consumer trust inference algorithm that can be relied upon by consumers in determining who to engage with in C2C e-commerce for specific contexts having not interacted directly with the seller/buyer in the physical world. There is a need by consumers to know whom they can trust on important C2C e-commerce transactions to limit their exposure to scams and fraudulent users on Twitter. This research sought to develop an algorithm to infer the trust score of a user engaged in consumer-to-consumer e-commerce using features present on his/her user profile. The algorithm utilized machine learning techniques. The algorithm provides consumers with a sense of trust in C2C engagements on Twitter. The research employed an experimental approach that involved the development of an algorithm and its validation. Wrapper approach was adopted for feature selection using data mined using Twitter Search API using C2C keyword-hashtag (#). Multi-class classification was successfully applied to infer a consumers trust score. Potential users can then use the proposed algorithm to check and choose trusted consumers on Twitter for different transactions.
An Automatic soiled linen detection prototype for hospital ward caregivers
(Strathmore University, 2019) Akumonyo, Lidonde James
The well-being of a hospital patient is highly prioritized by the provision of a clean and safe environment with the aim of improving the rate of patient’s recovery. Continuous patients’ checks and alert mechanisms are essential to the caregiver in aiding them to ensure a clean and safe environment is maintained for the patient receiving medical care from the facility. The existing system used relies on scheduled visits done by nurses during their shifts to assess the nature of the patients whilst assisting them where necessary. This process is usually cumbersome and prone to neglect which leads to patients getting new hospital-acquired infections. In other cases, the unpredictable nature of patient’s condition may result to high incontinence where the patient may soil linen more frequently due to deterioration of their health condition and this would need to have the caregiver promptly notified when such an event occurs to have the patient linen changed. With this challenge experienced, the study aims to come up with a solution termed as a soiled linen detection prototype to alert hospital ward caregivers of soiled linen. During this process, the researcher employed the use of experimental research to determine the variables essential in soiled linen detection. Together with this prototyping and use of questionnaires were employed to fine tune the system into meeting the users’ requirement as a solution to soiled linen detection challenge. The solution created integrates the idea of IoT with Wireless Sensor Networks. Hospital beds will be attached to humidity and gas sensors that are connected to the IoT device. The bedlinen will cover the sensors and they will transmit the hydrogen sulphide gas levels and humidity levels to the central cloud storage system whenever the safety levels have been exceeded. Upon successfully sending of the information from the IOT equipment the web application automatically picks data entry made to the cloud database storage and displays it on the client GUI awaiting to be handled by the caregiver. The client app GUI is used by the assigned caregiver and the nurse at the central ward station to notify and locate the bed that requires attention allowing proactive response to the patient who requires bedlinen change. The results of this was a prompt notification soiled linen prototype that was able to give notification as soon as the gas and humidity levels were exceeded
A Blockchain approach for detecting counterfeit academic certificates in Kenya
(Strathmore University, 2019) Otuya, Joy Atuwo
There has been an increase in the number of people seeking to pursue higher education and secure employment in different sectors across Kenya and the world at large. Consequently, there has been an exponential increase in the number of fake academic certificates being issued. Employers end up higher less qualified staff, while education institutions admit students with lesser qualifications. This has had a negative impact on education institutions in as far as credibility is concerned and for the employers, low productivity from the under qualified staff. Several methods have been put in place by different organizations for purposes of verifying academic certificates in Kenya. However, most of the methods employed are manual processes which are time consuming, tiresome and more prone to errors. The automated processes currently in use still have loopholes that can easily be exploited to bypass the verification process. For this reason, there is need to come up with an automated solution that will ensure proper verification of academic certificates thereby upholding the credibility of our academic institutions and increasing productivity in other organizations. This research explores different strategies employed in Kenya together with their challenges, the traditional and automated systems available for verification of academic certificates and finally proposed the development of a blockchain capable application for purposes of verifying academic certificates.
Dynamic vehicle routing model using geometric Brownian motion
(Strathmore University, 2019) Mungai, Joseph Muroki
Traffic jams are one of the most common problems associated with urban areas worldwide. The effects of traffic congestion are usually felt during peak hours of the day and in cases of accidents or bad weather. Essentially, Traffic jams occur with or without the aid of traffic Marshalls or traffic lights. However, the traffic jams get worse with absence of these agents especially in developing countries where public transport vehicles have little regard to traffic rules and regulations. As such, other modes of traffic avoidance and mitigation are required to mitigate the effects of traffic jams. A majority of city dwellers have resulted to waking up early in order to beat the traffic or leaving work early or later than usual. Therefore, to mitigate traffic jams there must be the provision of an alternative solution to the traffic menace. Response time by members of the police force would play a key role in reducing traffic congestion during accidents or incidents. The use of applications such as Waze and Google maps have greatly helped in navigation as well as informing the public on traffic situations on the roads. Radio stations and other social platforms like Twitter and Facebook have become a source of information with regards to traffic situations. Essentially, to substantially help reduce the effects of traffic jams in urban areas, more still needs to be done in terms of providing information to both road users and authorities with regards to traffic situations. Therefore, the need for alternative sources of information that aims at providing relevant information in regards to the traffic congestions. Notably, proper information through social media and other channels has led to a reduced response time by traffic Marshalls as well as a way to warn motorists to avoid areas with accidents as well as high traffic volumes. This research work proposes a dynamic vehicle routing model which implements geometric Brownian motion to help divert traffic to routes less congested. Notably, the model requires a large amount of data in order to accurately predict the expected traffic volume at a particular time. The system calculates the expected traffic at a particular time using variables calculated from the historical data of the route. It is therefore important for historical volume data to exist for the model to work. Based on the provided data, the system then calculates expected traffic volume for the routes entered by the user at a particular time and displays the same to the user. The system is able to calculate the expected traffic volume at a time (t) and is able to find an expected average of traffic on a route
Emergency response system based on an intelligent and optimal route finder
(Strathmore University, 2019) Ngunjiri, Kiere Peter
Road accidents have become a significant cause of injuries and death in developing countries. Every year, the lives of more than 1.25 million people around the world are cut short as a result of this road traffic crash menace. Approximately 20 to 50 million more people suffer non-fatal injuries, with which many incur disability as a result of their injury. Road traffic injuries cause considerable economic losses to individuals, their families, and the nations as a whole. These losses arise from the cost of treatment as well as lost productivity for those killed or disabled by their injuries, and for family members who need to take time off work or school to care for the injured. Road traffic crashes cost most countries 3% of their gross domestic product. Reports show that one of the best ways to reduce this fatalities and disabilities from these crashes is by decreasing the casualty evacuation time; this is the timely and efficient movement including en route care provided by medical personnel to injured patients evacuated from the scene of an accident to receiving medical facilities. This research aims at improving the evacuation process in order to reduce the fatalities as well as the impact of the injuries achieved by helping the public have access to specialized equipment and vehicles in aid of disaster and accident management. The purpose of this research has put more focus on ambulance access for patients in critical condition by designing a web-based program that can be accessed from a mobile phone or computer to inform the closest located ambulance to respond to distress calls. Also, there was need to help the ambulance drivers' get to the nearest hospital using the shortest and fastest route, which would be term as the most convenient route. The framework designed helps the patients who need emergency care attended to and their lives put out of danger as fast as possible for further medical attention. The framework also provides the emergency response unit with a platform to get fully reimbursed after delivery of services. An extensive literature review was carried out to determine the impact of delayed emergency response to patients in critical condition and addressed the problem through the development of a system that eliminates unnecessary delays.
A GPS-Based IoT system for Black Spot Notification
(Strathmore University, 2019) Owire, Billy
Accident black spot is a length of the road marked as having high road accidents potential. Most of Kenyan highways have become a nightmare to passengers, drivers and pedestrians in the recent past. The situation is traumatizing especially when traffic volumes are high. Most people have lost their lives while others left disabled. One of the common causes of these road accidents is a driver being unfamiliar with the roads. Most accidents have always occurred in the same place for example, in Kenya. Despite authorities using various methods to raise awareness of black spots including use speed guns detectors the rate of the accidents on the roads is still in an alarming rate. More often, signposts on the road can be vandalized or even be hit by a reckless driver. There is also an assumption that every driver reads and understands these signposts. The research proposed the development of an IoT system embedded on a vehicle and having ability to sense a black spot from a given distance using the global positioning system (GPS) coordinates. The system targets public transport vehicle since these vehicles get involved in road accidents on regular basis. After sensing the black spot, the system then provides an audio directive that notifies a driver to reduce the cruising speed and avoid overtaking. The proposed system was developed using GPS, microcontrollers, and GSM mobile technology. The system was tested using coordinates extracted from road sections marked us blackspots. The proposed system’s output was in form of notifications delivered to the user using the GSM system chip.
Intellibot data cleaner: a study of Kenya Revenue Authority’s data cleaning exercise
(Strathmore University, 2019) Odero, Jerry Omondi
Data cleaning is an activity involving detecting and correcting errors and inconsistencies in a database, data warehouse or any data record of an organization. Kenya Revenue Authority (KRA) in its quest to be a fully data driven organization, is actively undertaking the data cleaning process. However, this process is currently manual and slow as it involves physical transfer of documents to be processed from the various stations, via different levels of management for approval, to the centralized return processing unit. A process, which might take at least a fortnight for the processing of one taxpayer’s ledger account. Furthermore, this whole process needs lots of man-hours, since there is a vast amount of data to be cleaned due to the many ledger accounts affected during the manual filing system that ended in 2014. There exists many data cleaning processes and approaches which are used to purge out “dirty data”, before it’s loaded into the data warehouse. These processes vary depending on the data source, they are time consuming and expensive for organizations, in terms of skilled staff and the tools involved, hence this research proposed the application of RPA (Robotic Process Automation) to develop an intelligent bot (Intellibot) to be used in the transactional data cleaning exercise in Kenya Revenue Authority (KRA). With the transition from legacy system to I-Tax and I-CMS systems for domestic and customs revenue management respectively, the researcher sought to find out the current data cleaning process in the legacy system. This research led to the development of an RPA system for the current manual data cleaning process implemented and tested using the Blue Prism platform. The system detected the errors – using a knowledge-based model-, clustering them as errors due to uncaptured returns, uncaptured losses or credit re-adjustments. The intellibot system was able to load the ledgers, detect the errors and clean them with utmost precision. Experiments conducted on performance of the bots varied by seconds, in the first experiment. Also in the second performance test, there was a variance of seconds in cleaning the different errors detected, hence improving the data integrity significantly: free of errors, to be migrated to the I-Tax platform, thus support better decision making process in the organization, and a higher return on investments.
IoT Based realtime fish pond water quality monitoring model
(Strathmore University, 2019) Obado, Steve Agono
Fish farming contributes significantly to food security in developing countries throughout the world. It supplements income and is a source of protein in majority of rural and urban homes. Fish farmers however, are faced with numerous challenges during the production period and key among them is the monitoring and management of the production resources. Water is the main resource in the fish farming process. The lack of effective monitoring of water quality and timely intervention in instances where there are changes in the quality aspects such as temperature and water levels, has led to resource wastages and losses to the farmer due to low production. However, in Kenya most fish farmers are yet to adopt technology in their production processes. Most farmers rely on their experience when it comes to making decisions in control actions; which most of the time are susceptible to errors due to the complexities involved. The advent of the Internet of Things (IoT) is opening various ways in which farmers can rare fish by simply installing inexpensive sensors which are able to communicate with smartphones and provide a means of monitoring and managing the fish ponds. This research proposes a solution which is a real time fish pond water quality monitoring model which utilizes a smartphone that shall be adopted by the farmers. The model utilizes IoT concept which enables information gathering about water quality through the corresponding sensors. The status of the water quality aspects shall then be relayed on a real time basis through a cloud platform via a microprocessor to a graphical user interface (GUI) on the farmer’s smartphone. The farmer can then act as per the information relayed or the model can automatically act on behalf of the farmer as per the predefined actions. The data gathered by the model can be extracted in various forms for analysis. The real time information shall enable timely intervention by the farmers which eventually helps minimizing or eliminate wastages. A rapid prototyping methodology was used where the requirements of the system were gathered and the components designed and tested throughout the process.
A Location-based commodity ordering and delivery system: case of boarding schools in Nairobi
(Strathmore University, 2019) Otieno, Isaiah Ouma
There is general preference of boarding schools by parents in the bracket of middle and upper class to ease pressure of caring for the children vis a vis work pressure. The matter is compounded by new Kenyan government policies on employment of house helps and the distance of the schools from the location where the parents stay. From time to time, students in boarding schools require replenishing of simple supplies and consumable subject to school regulations. This scenario presents a new challenge in that families are unable to tend effectively to the needs of the students in boarding schools due to distance and time limitations. This research sought to explore the development of a system that allows online purchases and delivery of supplies to students in boarding schools by a touch of a button irrespective of the location of the school. The proposed solution consists of a Web-based and mobile application that allows a student to make a request that a parent approves before the delivery is made to the student. The system works by locating the supplier nearest to the school to reduce on the time taken to deliver the commodity as much as possible. The system utilizes a matrix that works by getting the nearest delivery person (Motorbike delivery person) that is closest to the supplier to save on time and cost. The supplier and the delivery person both get paid after successful delivery of the commodity Agile development methodology and object-oriented analysis and design were used in this. The final prototype of the location-based commodity ordering and delivery system was tested to ensure the functional and non –functional requirements were met by the system and the system was tested in fourteen schools in Nairobi County with eighty percent satisfaction rate from parents and students who used the system.
A Mixed Integer Programming optimization model for scheduling blood donors in disaster & emergency response: a case study of Nairobi region
(Strathmore University, 2019) Githogori, Samuel Maina
In recent years, Kenya has experienced tragedies ranging from natural disasters such as floods, terrorist activities such as the Westgate and Garissa University attacks, man-made tragedies such as road accidents and collapsed building, as well as tragedies resulting from reckless human behavior, such as fuel siphoning, and building next to power lines. When such kinds of disasters and tragedies occur, they have historically caused serious injuries that sometimes cause death. Also, during such events, medical emergencies arise, blood is one of the most critical components required by medial responder, and health facilities in order to perform transfusions that are necessary to save the lives of individuals. In the past, nationwide blood appeals have been conducted by authorities such as the Kenya Red Cross Society, media houses, politicians, and ordinary citizens, and Kenyans of Goodwill respond in large number at blood donation centers to donate blood. The challenge arising is that the system of appealing for blood is informal, unstructured and fragmented. It is difficult to track the effectiveness of ad-hoc methods of appealing for blood, and hard for potential blood donors to determine their eligibility in case they need to assist. The study proposed a mixed integer programming (MIP) model to optimize decision variables, which would determine the most optimal donation schedule and location for a given donor, based on whether they are eligible to donate, or not. The model sought to reduce the cost of responds, which is a function of the probability that a request for blood appeal will be posted, and the number of trips, distance, and cost it takes donor to respond. The model incorporated constraints such as donor availability within a given time block, and donor willingness to respond in a given region. The model’s outcome suggested that increased donor flexibility leads to a decrease in cost per donation session, and an increase in available regions increases donor flexibility, hence lower cost per donation intervention session on the donor.
Mobile application to assist cancer patients access healthcare and funding in Kenya
(Strathmore University, 2019) Chepchirchir, Serem Mercy
Cancer has posed to be a killer disease over time. Most patients die of cancer due to lack of access to early screening and effective treatment and lack of medical funds. Cancer treatment is deemed expensive thus most patients would need a health insurance fund or donor to offset the hospital bill. Cancer patients also need information on where they can access quality health service. There are several stakeholders who have joined forces to overcome the cancer burden. The government of Kenya has spearheaded the initiative by introduction of NHIF cover, covering cancer treatment. Cancer patients can be treated at NHIF approved hospitals using the capped cover amount. There are also Non-governmental organizations and cancer fund drives that have been organized to assist in the cancer burden in Kenya. Data was collected using questionnaires, the data collected was used to find out the challenges faced when accessing health care and funding. Data was analyzed using Google analytics tools and results obtained were used to guide the researcher in the design of the mobile and web application. Agile software methodology was used to develop the application. A mobile and web application was developed to assist cancer patients to access health care and funding. Users of the system are required to register their details that consist of their bio data as well as their NHIF number. The details of the user are sent to the donor to assist in paying the monthly contribution as well as the deficit amount to cover the full cancer treatment. Users are also referred to an oncologist and Hospital facility upon appointment request.
A Mobile application to track pension payments: a case for pensions department in Kenya
(Strathmore University, 2019) Odongo, Arthur
According to the existing terms and conditions of service especially for the permanent and pensionable officers in public service in Kenya, officers expect terminal benefits as an incentive for the services they rendered to the country for a number of years of their working life. These terminal benefits are referred to as pension in this study, which is simply the amount set aside either by employer, employee or both to ensure that at retirement there is something for employees to fall back on as an income so that they are not stranded financially. All public service retirees or pensioners are aware that they as entitled to some pension benefits. However, most pensioners do not know what their benefits are as well as not knowing how to access the benefits due to lack reliable, effective and efficient methods of interaction between them and the Pensions Department. This leads to suffering and hardship of pensioners and their dependents and beneficiaries of deceased public servants associated with delay in the processing and payment of pension. Current methods used for interacting with pensioners are through the media, sending letters or pensioners visiting the Pensions Department in Nairobi. These methods are costly, inefficient and unreliable. These challenges and weaknesses necessitated to the development a mobile application exclusively for the civil service pensioners whose pension is disbursed through the Pensions Department. The application helps pensioners interact directly with the Pensions Department by tracking their pension payments through an interactive menu based Unstructured Supplementary Service Data application that provides feedback and notifications through a Short Message Service. The study used V- shaped software development model to design, develop and test the application. The application was tested to show that pensioners can interact directly with the Department and access pension services using a mobile phone. Findings from usability testing illustrated that pensioners appreciated that the application was simple to use, provided clear feedback and instructions as well as being easy to learn, even for the first time. Overall, they were satisfied using the application and agreed to recommend it to other pensioners. Feedback from the respondents testing the application showed they liked the idea that such a digital tool could be accessed using any mobile phone without Internet connectivity or the need to download it. The application helped them familiarise themselves with the pensions processes by providing them with feedback or alerts as they tracked their payments till the payment stage. The study concluded by recommending the adoption of the application by the Pensions Department in Kenya and in other African countries with such public pension schemes.
A Mobile health application for patient monitoring in maternal healthcare
(Strathmore University, 2019) Chege, Ian Antony
With the recent global uptake of mobile devices and mobile connectivity services, a new avenue for the dispatch of health information services to patients by health facilities has emerged. Mobile health applications complementing patient monitoring systems have enabled access to health care services by patients whereby medical practitioners are able to monitor their patients remotely and therefore provide healthcare information either in real time or intermittently through the use of mobile devices. Of particular importance is the field of maternal health care, which has received a major boost by virtue of mobile penetration. This has brought about uptake of patient services and the receptiveness towards mobile capabilities for the patients to their respective health facilities. In Kenya, a majority of the number of women do not receive the recommended four or more required antenatal care visits. A contributing factor towards these low figures is that a number of these women live in abject poverty. In order to combat these low figures, monitoring of patients by doctors and community health workers through use of mobile applications will go a long way in increasing health care services to expectant mothers. This application area focuses on a mobile health application for gathering of patient data and dissemination of stage-based maternal health information to the patients’ mobile devices. The application also provides timely antenatal care tips as well as allowing the user to set reminders on upcoming appointments and expected antenatal care visits. The application also allows for two-way interactive text messaging between the expectant mother and the health worker. Testing and implementation of the application yielded 60% acceptability and user satisfaction from the respondents, with users determining the accuracy and responsiveness of the application at 62%. From the testing results provided by the respondents, the application can be deemed as usable and viable for implementation. In conclusion, the application was deemed acceptable by the respondents, who found the two-way interactive messaging functionality unique.
A Phishing detection model based on dynamic-hybrid feature selection
(Strathmore University, 2019-07) Ruiru, Daniel K
Phishing attacks have been a big internet nuisance since the early 1990s when hackers started stealing information from organizations using messaging platforms. At the time, the problem affected large institutions and corporations as the internet was still in its early stages of development and, had minimal individual subscribers. The early 2000s saw the widespread application of the technology (internet) which subsequently saw phishers target individual users using electronic mails (emails). In itself, phishing is a form of cyber-attack that steals personal information from unsuspecting users by duping them using verification or reward emails. This deception process ultimately helps the intruders to access sensitive data that can be used to access financial records for monetary gains or identity theft. Phishing attacks are so prevalent today that over 95 percent of all cyber-attacks are characterized by their intrusion procedures. Moreover, the attacks seem to increase each year and based on recent surveys are said to have a 60 percent annual growth rate. It is because of this outcome that this research proposes a predictive model to detect phishing attacks by implementing a system that pre-empts the intrusion processes before they happen. Unlike conventional methods that rely on human expertise to mitigate the problem, the proposed model automates the identification of the attacks and subsequently their control. This research aims to achieve this goal by optimizing the selection of subset features using a dynamic model that analyses the structural properties of phishing attacks to get adaptive attributes (features) for detecting phishing threats (as highlighted in chapter 4). Random forest is then used as the final classifier owing to its accuracy results (84.13%). Ultimately, the study then proposes the construction of a base model for bootstrapping other detection models in the cyber-security world.
A Real time electricity usage monitoring system using smart meters for wastage detection in Kenya
(Strathmore University, 2019) Obonyo, Joseph Oduor
Electricity utilities face dual challenges of generation and distribution. While many utilities are now migrating their operations into modern digitized platforms, most of these utilities have had to rely on old systems to carry out and monitor their commercial activities such as connectivity activity, itinerary scheduling, meter and device management, cycle billing, collections management and management reporting. This is causing utilities to reimagine customer engagements with a focus on feedback, put in place loss detection systems in their grids, use predictive models to schedule maintenance and other asset-management activities, equipping field workers with mobile devices that let them access technical instructions while in the field and deploy customized systems to help manage the extending networks. The main focus of this study was to design a real time electricity usage detection system using smart meters in Kenya, a system that would aid utility companies in providing real time feedback for service provision and anomaly detection. This research was done using smart meters to remotely record usage and record those messages manually into a web based portal. While it was not possible to replace existing old meters with the smart meters, the study made use of a test kit to simulate power consumption and usage recording. This study has put in place a model system that can be remotely connected to a smart meter and feedback generated in real time
Real-time monitoring system for drunk driver through Internet of Things - a case of Nairobi County, Kenya
(Strathmore University, 2019) Mbugua, Joseph Muoho
One of the major causes of traffic accidents and crashes globally is drunk driving. Though driving under influence of alcohol is illegal, drivers still find themselves doing it. This has resulted to deaths and fatalities which affect the economy negatively. Currently, different technologies have been adopted to reduce the vice without success. In Kenya, breathalyzers are being used by traffic police to monitor drunk drivers. This technology has failed due to corruption. The reason being the culprits are able to buy their way out. Technological innovation needs to be implemented in a cost effective, efficient and legal manner. This enables to combat drunken driving on the roads easily. The researcher applies the V-Model Methodology to design, implement and test of a realtime monitoring system for drunk driver through IoT. The system uses fuzzy logic based algorithm to analyze the response of MQ-3 and MQ-135 sensors. Sensor fusion is achieved through processing the analog to digital converted values of the sensor output using an algorithm to determine alcohol concentration in the breath (BAC). The analyzed result determines whether the Blood alcohol concentration (BAC) is within the legally permissible standards. Upon the detection of such a situation, an alarm is activated. Additionally, an ‘alert SMS’ indicating the drunk driver’s location as tracked by the GPS receiver on the same system and the vehicle registration number is communicated to the SACCO managers using GSM cellular network to take appropriate action of intercepting the vehicle. The tested real-time results indicated the successful implementation of the system.
Rogue access point detection framework on a multivendor access point WLAN
(Strathmore University, 2019) Barasa, Kunjira Fredrick
Wireless internet access has become common throughout the world. IEEE 802.11 Wireless fidelity (Wi-Fi) is now a common internet access standard almost becoming a requirement in homes, offices, universities and public places due to developments in Bring-Your-Own-Device (BYOD), mobile telephony and telecommuting. With the proliferation of Wi-Fi comes a number of information security challenges that have to be addressed. One of the major security threats that comes with Wi-Fi is the presence of rogue access points (APs) on the network. Unsuspecting employees in a company or attackers can introduce rogue APs to a secure wired network. The problem is amplified if the wireless local area network (WLAN) consist of multivendor APs. Malicious people can leverage on rogue APs to perform passive or active attacks on a computer network. Therefore, there is need for network administrators to accurately, with less effort, detect and control presence of rogue APs on multivendor WLANs. In this thesis, a solution that can accurately support detection of rogues APs on a multi-vendor AP WLAN without extra hardware or modification of AP firmware is presented. In the solution, information from beacon frames is compared to a set of approved parameters. Intervention of a network administrator is included to prevent MAC address spoofing. A structured methodology was adopted in developing the model on a Windows operating system. Python programming language was used in coding the system with Scapy and Tkinter as the main modules. SQLite database was used to store required data. The system was tested on a setup WLAN that composed of three different access points in a University lab. It was able to capture beacon frames sent by the access points and extracted MAC address, SSID and capability information as the key parameters used in identifying and classifying the access points. The system uses the captured information to automatically compare it against an existing database of authorized parameters. It is then able to classify an access point as either rogue or authorized. The system issued alerts that described the detected APs to a network administrator. The rest of this document gives details of scholarly works that are pertinent to the study, the research methodology used, implementation and testing of the model followed by discussions of findings and the conclusions and recommendations made by the researcher.
A Secure end to end verifiable e-voting system using cryptography: a case of Independent Electoral and Boundaries Commission of Kenya
(Strathmore University, 2019) Magonga, William
Elections that are run purely on manual voting system have presented a plethora of challenges that have often affected the credibility of elections done entirely using the traditional manual voting system. Countries round the world are quickly shifting to the use of e-voting systems with such nations already reaping the benefits of running their elections using e-voting systems. However, a number of countries especially in the developing world are yet to embrace a full e-voting platform and this has often affected the election outcomes in such countries. This study sought to evaluate the use of electronic voting systems to address the electoral malpractices in elections. The study proposes an e-voting system that employs a java enterprise application standard which makes use of the tiered design pattern allowing for scalability using distributed approach. All components are an extension of the server to the client machine, allowing for dynamic content delivery. Voters are only allowed to vote once. After voting, the user identity is flagged to prevent another attempt. Any further attempted logins will be denied access. Communication between the client machines and the servers are secured preventing eavesdropping or snooping on the user's identity in transit. Further once the data is delivered to the server, the application environment also provides added layer of security ensuring that only authorized personnel have permission to view data. To further enhance security all user actions on the client terminals are relayed to the server based on event response model. Security logs can be generated based on user activity which can only be viewed by users with administrative rights. The system uses two different databases. The electoral process that involves capturing data is managed using election database. The security aspect that involves starting/stopping the election, managing and viewing of security logs is managed using a security database. Both of them run on MySQL platform. This platform was used because of its compatibility with PHP and other platforms as well as the ability to encrypt passwords before storage. This ensures that confidentiality and integrity aspects are enforced.
Sentiment analysis for TV show popularity prediction: case of Nation Media Group’s NTV
(Strathmore University, 2019) Mutisya, Joshua Mutinda
Media-houses play a vital role in informing the masses on the issues that matter. They are also a source of entertainment for many households. In Kenya, the public depends on media largely for entertainment and educational purposes. However, many media-houses find it difficult to make decisions on what the viewers actually wish to watch. This makes the media-houses to be in the dark, unaware of what viewers want and making decisions based on perceptions rather than data. Most of the analytic tools used by media-houses in Kenya are used to provide insights on website-related activities such as site visits, number of article reads and read-depths. This is not a good way of measuring popularity and does not create a true reflection of how an audience perceives a given product. In this study, we propose a predictive model that uses sentiment analysis to determine the popularity of a given TV show. This enables accurate decisions to be made based on the viewership trends over a specific period of time. Natural Language Processing is used to perform sentiment analysis on tweets derived from Twitter. This solution involved tweets being derived from the social site Twitter through the use of the Twitter API. Once fetched, the tweets had their polarity measured through the use of a lexicon dictionary in order to remove neutral tweets. These tweets were then be labelled as either positive or negative using the Support Vector Machine classifier. Then the overall popularity score of a movie was calculated. The solution was able to not only show the polarity of derived tweets, but also assign overall popularity scores showing how positive or negative a TV show is.
Smart fertilizer recommendation through NPK analysis using Artificial Neural Networks
(Strathmore University, 2019) Siva, Faith
Agricultural practices, tools and technologies have taken a new paradigm. Precision agriculture is essential to ensure that site-specific crop management is implemented, which includes soil nutrient remedies per crop requirement. Though fertilization is key in boosting productivity, there is need for analysis of the potentials and limitations of soil as a basis of recommending the correct type, quantities and application time of fertilizers to counter uncertainty in fertilizer use. The complexity of finding the optimal fertilization range greatly contributes to major inefficiencies like productivity losses, resource wastage and increased environmental pollution caused by farmers’ use of intuition, trial and error, guesswork and estimation. With these, farmers cannot accurately predict what impact their decisions will have on the resulting crop yields and the environment. Some solutions implemented for soil fertility management such as use of mobile laboratories or imported equipment have had their share of challenges such cost of implementation, ease of use and adaptation to the local environment. Other available solutions including taking soil to laboratories for testing is tedious, time consuming and inconsistent. This study proposed development of an ANN model that predicts NPK nutrient levels and recommends the best fertilizer remedy and application time based on the weather forecast. This involved use of IoT, machine learning techniques and a weather API through RAD methodology and experimental research design. Historical data of temperature, PH and NPK from KALRO Library was used to train and validate the model. The developed model achieved an RMSE 0.5 with 75% of data used for training and 25% used for testing. This is in effort to encourage precise fertilizer production for particular areas of need.

Browse

Browsing MSIT Theses and Dissertations (2019) by Title

Results Per Page

Sort Options