2,234 research outputs found
Tor over QUIC
No full textTor is the most popular tool for anonymous online communication. However, the performance of Tor's volunteer-run network is suboptimal when network congestion occurs. Within Tor, many connections are multiplexed over a single TCP connection between relays, which causes a head-of-line blocking problem, degrading relay performance. In this thesis, Tor's TCP transport layer protocol is replaced by QUIC, a UDP-based protocol that natively supports multiplexing streams asynchronously, effectively solving head-of-line blocking. Its performance is evaluated within various network environments through Containernet, a flexible Docker-based network test bed that allows for simple reproduction of results. Along with testing multiple congestion control algorithms, the impact of using Hystart++ within Tor over QUIC is evaluated. It is found that QUIC over Tor can perform up to 50% better in time to last byte performance than vanilla Tor in a realistic network environment, while featuring more consistent time to first byte performance. Additionally, the evaluations shows that throughput consistency and fairness amongst downloaders are improved as well, Besides offering improved performance, Tor over QUIC is designed with deployability and security in mind. This makes QUIC an attractive replacement as Tor's transport layer protol.Computer Scienc
Omprövning av kundvärde : En analys av RFM-baserad kundsegmentering och lönsamhet
No full textCustomer segmentation is the process of grouping customers into distinct groups to enable the development of targeted marketing and customer relationship management strategies. RFM is a prominent segmentation method which identifies valuable customers based on engagement and spending. However, the cost of serving customers is often overlooked, which, in the context of subscriptions, becomes problematic as increased engagement drives costs while revenues remain static. Thus, valuable customers, as viewed through the lens of RFM, are not always profitable. This study addresses this research gap by comparing the RFM-based values of customer segments with their profitability, as measured by gross profit and gross profit margins. A data-driven approach was employed, utilising RFM and k-means clustering in a car wash context. In total, four subscriber segments and five pay-per-use segments were generated. The comparison shows that RFM-based customer values do not always align with profitability. High-engagement customers risk reducing profitability, while low-usage customers often deliver better margins than their RFM scores imply. This discrepancy was especially relevant for subscribers, where frequent usage under a flat fee eroded margins. These findings underscore the importance of incorporating cost-to-serve into segmentation analysis to identify valuable customers more accurately. The study extends segmentation approaches by integrating profitability considerations into RFM analysis while applying the methods in a novel context. Furthermore, it informs managers to evaluate and prioritise customer segments based on other parameters than engagement and spending.Kundsegmentering innebär att gruppera kunder i distinkta grupper för att möjliggöra utvecklingen av riktade strategier för marknadsföring och kundrelationer. RFM är en framträdande segmenteringsmetod som segmenterar kunder baserat på tre beteendebaserade egenskaper och identifierar värdefulla kunder baserat på nivån av engagemang och spenderande. Dock förbises i många fall kostanden för att betjäna kunder. Vilket i abonnemangssammanhang blir problematiskt då ökat engagemang driver kostnader medan intäkterna är statiska. Därmed behöver inte värdefulla kunder enligt RFM vara detsamma som lönsamma kunder. Den här studien ämnar åtgärda denna kunskapslucka i segmenteringsforskningen genom att jämföra RFM-baserade kundvärden med deras lönsamhet, mätt som bruttovinst och bruttovinstmarginaler. En datadriven metod bestående av k-means klustring och RFM tillämpades i ett biltvättssammanhang. Totalt genererades fyra abonnentsegment och fem segment för engångskunder. Jämförelsen visar att RFMbaserade kundvärden inte alltid överensstämmer med lönsamhet. Kunder med högt engagemang riskerar att minska lönsamhet medan kunder med lågt engagemang kan uppvisa bättre marginaler. Denna skillnad mellan RFM och lönsamhet var särskilt tydlig för abonnenter där frekvent användning inte har någon påverkan på intäkter men ledde till högre kostnader. Detta understryker vikten av att inkludera kostnader i segmenteringsanalyser för att identifiera verkligt värdefulla kunder mer träffsäkert. Studien vidareutvecklar segmenteringsmetoder genom att integrera lönsamhetsaspekter i RFM-analysen. Tilläggsvis informerar den beslutsfattare att utvärdera och prioritera kundsegment baserat på andra faktorer än engagemang
Omprövning av kundvärde : En analys av RFM-baserad kundsegmentering och lönsamhet
No full textCustomer segmentation is the process of grouping customers into distinct groups to enable the development of targeted marketing and customer relationship management strategies. RFM is a prominent segmentation method which identifies valuable customers based on engagement and spending. However, the cost of serving customers is often overlooked, which, in the context of subscriptions, becomes problematic as increased engagement drives costs while revenues remain static. Thus, valuable customers, as viewed through the lens of RFM, are not always profitable. This study addresses this research gap by comparing the RFM-based values of customer segments with their profitability, as measured by gross profit and gross profit margins. A data-driven approach was employed, utilising RFM and k-means clustering in a car wash context. In total, four subscriber segments and five pay-per-use segments were generated. The comparison shows that RFM-based customer values do not always align with profitability. High-engagement customers risk reducing profitability, while low-usage customers often deliver better margins than their RFM scores imply. This discrepancy was especially relevant for subscribers, where frequent usage under a flat fee eroded margins. These findings underscore the importance of incorporating cost-to-serve into segmentation analysis to identify valuable customers more accurately. The study extends segmentation approaches by integrating profitability considerations into RFM analysis while applying the methods in a novel context. Furthermore, it informs managers to evaluate and prioritise customer segments based on other parameters than engagement and spending.Kundsegmentering innebär att gruppera kunder i distinkta grupper för att möjliggöra utvecklingen av riktade strategier för marknadsföring och kundrelationer. RFM är en framträdande segmenteringsmetod som segmenterar kunder baserat på tre beteendebaserade egenskaper och identifierar värdefulla kunder baserat på nivån av engagemang och spenderande. Dock förbises i många fall kostanden för att betjäna kunder. Vilket i abonnemangssammanhang blir problematiskt då ökat engagemang driver kostnader medan intäkterna är statiska. Därmed behöver inte värdefulla kunder enligt RFM vara detsamma som lönsamma kunder. Den här studien ämnar åtgärda denna kunskapslucka i segmenteringsforskningen genom att jämföra RFM-baserade kundvärden med deras lönsamhet, mätt som bruttovinst och bruttovinstmarginaler. En datadriven metod bestående av k-means klustring och RFM tillämpades i ett biltvättssammanhang. Totalt genererades fyra abonnentsegment och fem segment för engångskunder. Jämförelsen visar att RFMbaserade kundvärden inte alltid överensstämmer med lönsamhet. Kunder med högt engagemang riskerar att minska lönsamhet medan kunder med lågt engagemang kan uppvisa bättre marginaler. Denna skillnad mellan RFM och lönsamhet var särskilt tydlig för abonnenter där frekvent användning inte har någon påverkan på intäkter men ledde till högre kostnader. Detta understryker vikten av att inkludera kostnader i segmenteringsanalyser för att identifiera verkligt värdefulla kunder mer träffsäkert. Studien vidareutvecklar segmenteringsmetoder genom att integrera lönsamhetsaspekter i RFM-analysen. Tilläggsvis informerar den beslutsfattare att utvärdera och prioritera kundsegment baserat på andra faktorer än engagemang
TOR Is required for the retrograde regulation of synaptic homeostasis at the drosophila neuromuscular junction
No full textHomeostatic mechanisms operate to stabilize synaptic function; however, we know little about how they are regulated. Exploiting Drosophila genetics, we have uncovered a critical role for the target of rapamycin (TOR) in the regulation of synaptic homeostasis at the Drosophila larval neuromuscular junction. Loss of postsynaptic TOR disrupts a retrograde compensatory enhancement in neurotransmitter release that is normally triggered by a reduction in postsynaptic glutamate receptor activity. Moreover, postsynaptic overexpression of TOR or a phosphomimetic form of S6 ribosomal protein kinase, a common target of TOR, can trigger a strong retrograde increase in neurotransmitter release. Interestingly, heterozygosity for eIF4E, a critical component of the cap-binding protein complex, blocks the retrograde signal in all these cases. Our findings suggest that cap-dependent translation under the control of TOR plays a critical role in establishing the activity dependent homeostatic response at the NMJ
Adding QUIC support to the Tor network
No full textPrivacy in the Internet is under attack by governments and companies indiscriminately spying on everyone. The anonymity network Tor is a solution to restore some privacy, however, Tor is slow in both bandwidth and latency. It uses a TCP-based connection to multiplex different circuits between nodes and this causes different independent circuits to interfere with each other. To solve this, we propose a transport layer implementation using the UDP-based protocol QUIC, as it allows independent streams over a single connection. We built a Tor prototype that uses this protocol and evaluated its performance using a custom network simulator, as existing simulators were shown to be incompatible. We show that the QUIC-based implementation increased performance in several of the use case scenarios, mainly outperforming on the ‘time to first byte’ metric.Electrical Engineering | Embedded System
Measuring accessibility of popular websites while using Tor
No full textTor is an anonymity network used by a vast number of users in order to protect their privacy on the internet. It should not come as a surprise that this service is also used for abuse such as Denial of service attacks and other malicious activities because of the anonymity it provides. For protecting themselves from this abuse, websites block Tor in various ways. We investigate the extent and frequency of this kind of blocking by requesting the Alexa top 1000 websites with and without Tor with the objective of highlighting the differential treatment observed by privacy-minded users. We build upon existing studies by using diverse metrics to measure discrimination and by extending our search to three sub pages of websites for detecting sophisticated blocking. We find at least 25.8% of the Alexa top 1000 websites discriminating on the home page against Tor users as opposed to 20.03% observed in previous studies. This number rises to 31.7% after including the three sub pages. We also discover new types of blocks such as Tor users being served old or different versions of websites. We categorize the blocked websites and find that Online Shopping and Finance/ Banking categories discriminate most against Tor while Social Networking sites and Search Engines discriminate the least.CSE3000 Research ProjectComputer Science and Engineerin
Products on Tor
No full textIn 1974 work establishing the collapse of certain Eilenberg-Moore spectral
sequences, Munkholm constructs, in passing, a bilinear multiplication operation
on Tor of a triple of -algebras. In 2020, the present author,
pursuing a multiplicative collapse result extending Munkholm's, studied a
variant of this product, without actually showing it agrees with Munkholm's. In
2019, Franz had defined a weak product on the two-sided bar construction of a
triple of -algebras under similar hypotheses, with which this author
proved a related collapse result, but without investigating the properties of
the induced product on Tor.
The present work demonstrates that the two products on Tor agree and are
induced by the product of Franz.Comment: 19 pages, comments welcom
Bestämma Betydande Attribut för Klassificering av Melanom Genom Attributselektion
No full textSkin cancer is a common disease and malignant melanoma is the most dangerous form of it. Although dangerous, the survival rate of melanoma patients is high if the diagnosis is made at an early stage. Computer aided diagnostics has been shown to have potential in accurately diagnosing the disease utilizing machine learning. Thus, machine learning algorithms can be used to effectively classify a skin lesions as either benign or malignant. These algorithms can be made more accurate and efficient by applying feature selection since it decreases the dimensionality of the feature space. The aim of this study is to apply feature selection on four different classifiers to compare morphological and SIFT features in order to determine which features are important for classifying melanoma. The results show that morphological features in general had a higher importance than the SIFT features, although this varied between different classifiers. Furthermore, forward selection was more effective than backward selection in terms of accuracy for three out of the four classifiers. Lastly, two morphological features were significantly more important than the other features. The most effective feature measured the compactness of the lesion and the second most described the contrast between the lesion and the surrounding skin in terms of the color red.Hudcancer är en vanlig sjukdom och malignt melanom är den farligaste formen av hudcancer. Trots att formen är farlig så är sannolikheten att en patient kan botas från sjukdomen hög om diagnosen sker i ett tidigt stadium. Datordriven diagnostisering har visat sig kunna effektivt diagnostisera sjukdomen med hög säkerhet genom att tillämpa maskininlärning. På så vis kan maskininlärningsalgoritmer användas för att klassificera hudutslag som godartade eller inte. Dessa algoritmer kan effektiviseras genom att utföra attributurvalsmetoder då det minskar antalet dimensioner som behöver beräknas. Syftet med denna studie är att undersöka vilka attribut som är viktiga för klassificeringen. Detta gjordes genom att tillämpa attributurvalsmetoderna Sekventiell Framåt- och Bakåtselektion på fyra olika maskininlärningsalgoritmer med indata i form av morfologiska och SIFT-attribut. Resultaten visar att morfologiska attribut generellt föredrogs i större utsträckning än SIFT-attribut, detta varierade dock mellan olika klassificeringsmodeller. Vidare var Framåtselektion mer effektiv än Bakåtselektion sett till träffsäkerhet för tre av klassificeringsmodellerna. Slutligen var två attribut mer effektiva än de andra. Det mest effektiva attributet beskrev utslagens kompakthet och den andra beskrev kontrasten mellan utslagen och huden runtomkring med avseende på den röda färgen
Android Tor Tribler Tunneling (AT3): TI3800 Bachelorproject
No full textTribler is a decentralized peer-to-peer file sharing system. Recently the Tribler development team has introduced anonymous internet communication using a Tor-like protocol in their trial version. The goal of our bachelor project is to port this technology to Android devices. This is a challenging task because cross-compiling the necessary libraries to the ARM CPU architecture is uncharted territory. We have successfully ported all dependencies of Tribler to Android. An application called Android Tor Tribler Tunneling (AT3) has been developed that tests whether these libraries work. This application downloads a test torrent and measures information such as CPU usage and download speed. Based on this information we have concluded that it is currently not viable to run the anonymous tunnels on an Android smartphone. Creating circuits with several hops that use encryption is very computationally expensive and modern smartphones can hardly keep up. By using optimized cryptographic libraries such as gmp or with the recently announced ARMv8 architecture which supports hardware-accelerated AES encryption, creating such circuits might become possible.Tribler developmentParallel and Distributed Systems groupElectrical Engineering, Mathematics and Computer Scienc
Bestämma Betydande Attribut för Klassificering av Melanom Genom Attributselektion
No full textSkin cancer is a common disease and malignant melanoma is the most dangerous form of it. Although dangerous, the survival rate of melanoma patients is high if the diagnosis is made at an early stage. Computer aided diagnostics has been shown to have potential in accurately diagnosing the disease utilizing machine learning. Thus, machine learning algorithms can be used to effectively classify a skin lesions as either benign or malignant. These algorithms can be made more accurate and efficient by applying feature selection since it decreases the dimensionality of the feature space. The aim of this study is to apply feature selection on four different classifiers to compare morphological and SIFT features in order to determine which features are important for classifying melanoma. The results show that morphological features in general had a higher importance than the SIFT features, although this varied between different classifiers. Furthermore, forward selection was more effective than backward selection in terms of accuracy for three out of the four classifiers. Lastly, two morphological features were significantly more important than the other features. The most effective feature measured the compactness of the lesion and the second most described the contrast between the lesion and the surrounding skin in terms of the color red.Hudcancer är en vanlig sjukdom och malignt melanom är den farligaste formen av hudcancer. Trots att formen är farlig så är sannolikheten att en patient kan botas från sjukdomen hög om diagnosen sker i ett tidigt stadium. Datordriven diagnostisering har visat sig kunna effektivt diagnostisera sjukdomen med hög säkerhet genom att tillämpa maskininlärning. På så vis kan maskininlärningsalgoritmer användas för att klassificera hudutslag som godartade eller inte. Dessa algoritmer kan effektiviseras genom att utföra attributurvalsmetoder då det minskar antalet dimensioner som behöver beräknas. Syftet med denna studie är att undersöka vilka attribut som är viktiga för klassificeringen. Detta gjordes genom att tillämpa attributurvalsmetoderna Sekventiell Framåt- och Bakåtselektion på fyra olika maskininlärningsalgoritmer med indata i form av morfologiska och SIFT-attribut. Resultaten visar att morfologiska attribut generellt föredrogs i större utsträckning än SIFT-attribut, detta varierade dock mellan olika klassificeringsmodeller. Vidare var Framåtselektion mer effektiv än Bakåtselektion sett till träffsäkerhet för tre av klassificeringsmodellerna. Slutligen var två attribut mer effektiva än de andra. Det mest effektiva attributet beskrev utslagens kompakthet och den andra beskrev kontrasten mellan utslagen och huden runtomkring med avseende på den röda färgen
- …
