No abstract

No abstract

The user interface for inputting text-based passwords has been the same for past 40 years. Today, technology enables the development of intuitive, interactive and responsive interfaces that can help users in creating and remembering more secure passwords. In this paper, we exploit the power of modern-day technologies and develop two novel interfaces, (i) linear one called as Pass-Scroll and (ii) circular one called as Pass-Roll. These graphical interfaces allow users to perform rotation operation by choosing a new starting point for their passwords. Consequently, the security of a n length password can be potentially improved by log2(n) bits. To evaluate Pass-Roll and Pass-Scroll interfaces we conduct two user studies, one in the laboratory and the other on Crowd-Flower. Both studies show that users willingly take advantage of these interfaces and choose a new starting point to rotate their password. We find that users' choices are quite diverse and multiple cues associated with the interfaces help users to recall their starting point in just one attempt. Moreover, our interfaces require no server-side changes and can be easily implemented as browser extensions. Permission to freely reproduce all or part of this paper for noncommercial purposes is granted provided that copies bear this notice and the full citation on the first page. Reproduction for commercial purposes is strictly prohibited without the prior written consent of the Internet Society, the first-named author (for reproduction of an entire paper only), and the author's employer if the paper was prepared within the scope of employment.

Few-shot learning aims to tackle the limitations, of classical techniques, of needing large data sets for learning an effective model. Generally, in few shot learning we have a meta-learning stage (learning to learn) which shall be followed by the few examples to learn upon. Considered as a hallmark of human intelligence, the community has recently witnessed several contributions on this topic, in particular through meta-learning, where a model learns how to learn an effective model for fewshot learning. The main idea is to acquire prior knowledge from a set of training tasks, which is then used to perform (few-shot) test tasks. Most work assumes that both training and test tasks are drawn from the same distribution, and a large amount of labelled data is available in the training tasks. This is a very strong assumption which restricts the usage of meta-learning strategies in the real world where ample training tasks following the same distribution as test tasks may not be available. In this paper we shall see a possible approach for the above problem. In this approach, we use Model Agnostic MetaLearning by Finn et al. to address the problem of few shot learning and an adversarial approach to tackle to the problem of domain shift in the target dataset.

Online Social Networks (OSNs) such as Twitter and Facebook have become popular communication and information sharing tools for hundreds of millions of individuals in recent years. OSNs not only make peoples life more connected, but also attract the interest of spammers. Twitter spam generally contains deceptive information, such as free voucher and weight loss advertisement to attract the interest of victims. A comprehensive analysis on the deceptive information will be of great benefit to the detection of Twitter spam. Twitter has now become one of the largest online social network sites. Over 500 million registered users spend vast time making friends with people who they are familiar with or interested in. For Twitter users, after relationships are built, they can receive tweets, usually something interesting or recent activities shared by their friends. Nowadays, Twitter has largely shortened the distance of people, and reshaped the way they communicate with each other. Current security experts suggest the best defense against spam is to educate Internet users to never click suspicious links in tweets. In the real world, however, the effectiveness of education is far from our expectations. Spammers leverage some certain deceptive topics, such as gain followers, cracked games, i.e. to lure users to click their malicious links. We refer this kind of information as deceptive information. The deceptive information is one of the key factors to the spreading efficiency of spam on Twitter. A better understanding of deceptive information is crucial to spam detection techniques. Therefore, we are motivated to thoroughly study the deceptive information employed by spammers.