CPMap: Design of Click-Points Map-Based Graphical Password Authentication

,


Introduction
Nowadays, traditional textual passwords are still the most widely deployed user authentication method across many organizations, but they have well-known limitations in the aspects of both security and usability [35].For instance, users are difficult to remember a complicated password for a long period of time; thus, users may select weak passwords for better recall.Generally, a weak password can be easily guessed by attackers, which would greatly degrade the security of authentication.In practice, this situation would be even worse than we previously believed (i.e., most created passwords only provided fewer than 10 bits of security against an online trawling attack) [1,34].
In some early studies, people were found to generally remember and recognize images better than textual passwords [24,26].Based on this observation, graphical passwords (GPs) are developed as a promising alternative to complement textual password-based authentication, which typically require users to create their credentials on images, i.e., DAS, PassPoints and CCP are some well-known schemes.In particular, DAS was proposed by Jermyn et al. [11], which allows users to draw their passwords on a 2D grid.Wiedenbeck et al. [33] developed PassPoints that demands users to generate passwords by clicking on any place on an image.Chiasson et al. [2] proposed Cued Click Points (CCP), in which the next image displayed can be varied with the previous click-point and users have to select five points in a sequence of images.
In order to obtain a large password space, map-based GP schemes are developed that can provide more potential places for users by adopting a world image.PassMap [30] and GeoPass [32] are two typical examples: PassMap asks users to choose two places in a sequence at any zoom-level on a world map, while GeoPass only needs users to select one location on a world map at zoom level of 16.Intuitively, the selection of one location is more vulnerable to shoulder surfing attacks, while increasing the number of locations may cause more additional burden on users (i.e., selecting two places).Previous study [23] investigated this issue and found that increasing the number of locations from one to two would not much degrade the performance of users' memorability, but would indeed consume more authentication time.
Contributions.In this work, we first conduct a user study to investigate the password patterns of PassMap, i.e., measuring the distance between the selected locations.Then, we design a new type of map-based passwords by combining existing geographical authentication with click points.We further conduct another user study to explore its performance as compared to PassMap and GeoPass.The contributions of this work can be summarized as follows.
-In the first study, we investigate how users would select the two locations on a world map under PassMap.The selection of PassMap is due to its scheme design and popularity.It is identified that common users are likely to create two locations that are very close to each other for the sake of time consumption during authentication (i.e., two locations are within the same community, which may greatly reduce the effective password space).-To enhance the existing map-based GPs, we develop a click-points map-based password scheme, named CPMap.Users have to firstly select one location on a world map and then click one point or an object on an image that is relevant to the previously selected location.This scheme can be regarded as a combination of geographical passwords and CCP.-To explore the performance of CPMap, we conducted another user study with 50 participants and compared our scheme with PassMap and GeoPass.Ex-perimental results indicate that our scheme can achieve better performance in terms of both security and usability.
Road map.The remaining parts of this paper are organized as follows.In Section 2, we introduce related studies regarding existing GP schemes, especially map-based authentication schemes.Section 3 describes our first user study on PassMap with 30 participants.In Section 4, we detail our proposed CPMap and analyze the results obtained from another user study with 50 participants.Finally, we conclude our work with future directions in Section 5.

Related Work
This section introduces the classification of graphical passwords and related research on map-based authentication schemes.
-Recognition-based scheme.This kind of schemes asks users to select one or more images from an image pool.PassFaces [25] is one particular recognitionbased scheme, which needs users to identify several human faces for authentication.Another scheme, called Story [5], requires users to pick out some assigned images from an image pool such as people, food, fruit, etc. -Pure recall-based scheme.This type of schemes requires users to create a pattern on an image as their secret.For example, Jermyn et al. [11] proposed a scheme of DAS ('draw-a-secret') that allows users to create their secrets on a grid.Tao et al. [31] introduced Pass-Go that asks users to create a password by selecting intersections on a grid.Based on the idea of Pass-Go, Android unlock patterns have been widely adopted on Android phones, allowing users to unlock the phone if they can input a correct pattern. 6everal other relevant schemes can be referred to [7,12].-Cued recall-based scheme.This kind of schemes needs users to select a sequence of points on an image or multiple images to construct their passwords.A typical system of PassPoints was proposed by Wiedenbeck et al. [33], which requires users to remember a sequence of five points on different images.To improve the memorability of PassPoints, Chiasson et al. [4] then proposed Persuasive Cued Click-Points (PCCP), which requires users to select a point on each of a sequence of background images.
To further enlarge the password space, a set of hybrid GP schemes were also developed in the literature, like click-draw based GP scheme [14] that combined the main input types of current GPs including clicking, selecting and drawing.Some other relevant studies on GP improvement can be referred but not limited to [6, 10, 15-17, 19-21, 36].
Map-based Graphical Password Schemes.To the best of our knowledge, Fox in 2010 [8] first presented the idea of using a digital map to create a password.After that, Spitzer et al. [27] developed a scheme called Cued Click Points (CCP ), which could combine the graphical approach with user's familiarity with navigating through Google maps.For implementation, users were presented with an image of the United States and simply click to where the key destination is located.Their results with around 50 participants indicated that around 60% users rated the system as easier to remember than traditional textual passwords in terms of memorability.
Map-based GP schemes become more popular from the year of 2012.Georgakakis et al. [9] proposed a scheme called NAVI, which allows users to draw a route on a pre-loaded map image.They initially analyzed the strength of the password, but did not give any user study for the real performance.Then, Sun et al. [30] proposed a map-based authentication system called PassMap, which allows users to select two sequenced places on a world map.In the evaluation, users found that PassMap passwords are easier to remember than textual strings.Similarly, Thorpe et al. [32] proposed GeoPass, a digital map-based GP scheme, which allows users to choose only one place on a world map as the credentials.Then, MacRae et al. [13] proposed GeoPassNotes, asking users to further select a note associated with their chosen location in the second step.Shin et al. [28] further implemented a modified version of GeoPass on a mobile device.The major difference between PassMap and GeoPass is the number of locations allowed by the system, i.e., clicking one or two places on a world map.Focused on this issue, Meng et al. [23] conducted a study with 60 participants and found that participants could perform very closely for both schemes.In other words, there is no significant difference between the selection of one or two locations.
To enhance the performance of multiple password memory, Meng [18] proposed RouteMap, a map-and route-based graphical password scheme, allowing users to draw a route on a world map as their secrets.In the user study with 60 participants, it is found that RouteMap can outperform similar schemes.Then, Meng et al. [22] conducted a study with 60 participants to investigate the recall of multiple passwords between text passwords and map-based passwords under various account scenarios.In particular, each participant has to create six distinct passwords for different account scenarios.It is found that participants in the map-based graphical password scheme could perform better than the textual password scheme in both short-term (one-hour session) and long term (after two weeks) password memorability tests.

A Study on PassMap
In this section, we conduct a user study with 30 participants to investigate the password patterns of PassMap.The selection is due to its scheme design and popularity.As introduced earlier, this scheme requires users to select two places on a world map at any zoom level.For authentication, users have to select the same location in the correct sequence and zoom level.PassMap Implementation.In this work, we adopted an open-source GP platform from the previous study [23].It enables an extensive move-by-dragging, zooming and search functions by leveraging the JavaScript from Google Maps API.The search function allows users to shift to a specific part of the map quickly and further locate a specific area.As shown in Fig. 1 (a) and Fig. 1 (b), users can input their usernames, and zoom in or zoom out on the map to find a place with zoom levels.According to PassMap [30], our system embedded a 640 × 420 pixel frame block for displaying the world map in a web page and road/map view was implemented by default with a tolerance of 21 × 21 pixels.
User Study.To investigate the password patterns of PassMap, we performed a study with 30 participants, who are volunteers and have not attended any courses in relation to security.The information of participants is shown in Table 1.
Before the study, we introduced our objectives to all participants and provided a guideline including all steps in the lab study.Each participant has 3 trails to get familiar with PassMap system.In particular, every participant was Result analysis.In the study, our major purpose is to investigate the password patterns created by participants, especially the distance between two selected locations.This is because two similar points may greatly reduce the effectiveness of PassMap (i.e., resulting in a weak password).To facilitate the illustration, we classify the patterns into different categories based on the in-between distance: community, district, city, nation (same continent), nation (different continents).
In total, we collected 90 PassMap passwords in the study.The distribution of in-between distance is depicted in Fig 2.
It is known that GeoPass is vulnerable to offline guessing attacks due to the selection of only one location.Intuitively, selecting two locations would double the password space of PassMap as compared to GeoPass.However, it is found that up to 37 and 22 PassMap passwords (65.6%) dropped into the first two categories.The close locations could cause PassMap to offer weak security against online guessing attacks, if the attacker had some effectively prioritizing guesses.In other words, choosing two close points would greatly lower the effective password space and increase the cracking probability.
To find the reason, we informally interview the participants and identified that most of them would choose two close locations to reduce the time consumption.for password creation, confirmation and login, respectively.In the interview, most participants reflected that it is very time-consuming to complete a successful authentication under PassMap; thus, they decided to select two close locations to reduce the time in zooming the map.
Feedback and Discussion.According to the results obtained in the study, it is found that most participants did not create a strong PassMap password.Table 3 further analyzed the feedback forms collected in the study.Ten-point Likert scales were used in each feedback question, where 1-score indicates strong disagreement and 10-score indicates strong agreement.It is visible that most participants still gave positive feedback on password creation (with a score of 8.2), but they believed the login phase should be improved, especially the time consumption was unacceptable.By considering the usability, most of them believed that current PassMap was not mature enough for real-world usage.
As users have to find two locations on a world map, PassMap can be considered as a two-step scheme.For such kind of schemes, there is a need to enhance the password design, especially to reduce the time consumption for better authentication in practical applications.

CPMap and Evaluation
To design an appropriate two-step geographical password scheme, a balance should be made between location number and time consumption.For instance, PassMap requires users to choose two locations on a world map, which may consume a lot of time during authentication.Focused on this issue, we design CPMap, a click-points map-based GP by combining the geographical passwords with click-point schemes.We then conduct a user study to investigate its performance as compared with PassMap and GeoPass.

Design of CPMap
To improve the performance of GeoPass, MacRae et al. [13] proposed GeoPass-Notes, which requires users to choose a note to be associated with their chosen location in the second step.They actually combined the location password with a note, where users are authenticated by correctly entering both a location and an annotation.However, we notice that writing a note may increase the time consumption and be less convenient on some mobile devices with a small touchscreen.In this work, we design CPMap, which is a combination of map-based password and click-points.More specifically, it needs users to select one location on a world map at first, and then click one point or an object on an image that is extracted from the surroundings related to the selected location.
Our scheme.Fig. 3 (a) details the steps on how to construct a CPMap password, and Fig. 3 (b) shows a concrete example of password creation.Firstly, similar to most map-based schemes, users have to select one location on a world map, e.g., the central station in Copenhagen.Then, CPMap shows up an image that is related to the selected location, e.g., a construction work environment in the central station.Similar to click-based GPs, users have to click one point (or an object) to create their password, i.e., selecting a tower (upper right corner).
For authentication, users have to locate the right place on the world map and select the correct object.In short, CPMap is believed to have several advantages over PassMap and GeoPass.
-CPMap allows users to select only one location on a world map, which aims to reduce the time for zooming out / zooming in on the map to find another location, as compared to PassMap.In our first study, PassMap was found to cost too much time for password creation and login.-The password space of PassMap is expected to reach around 2 36.9×2 with two map locations, whereas users are likely to create a weak password, resulting in a similar password space to GeoPass (2 36.9 ).By contrast, CPMap is also a two-step scheme: it requires users to decide a location at first while selecting an object on an image in the second step.As a result, CPMap can offer a larger password space than GeoPass.-Based on the previous studies on map-based password schemes [13,23,32], users were often advised to choose a familiar location where they have travelled or visited before.As CPMap provides an image that is related to the selected location, it can facilitate users to remember the created location.
CPMap Implementation.A prototype system of CPMap was implemented in our lab environment.Similar to PassMap, we use Java scripts and Google Maps API to fetch a real world map, in which users can perform a set of actions such as move (drag), zoom in, zoom out and search.Based on the Google Maps API, a particular surrounding image could be extracted, which can vary with the selected locations.We set the error tolerance to a 21 × 21 pixel box around the place.As a comparison, it is worth noting that GeoPass and PassMap has an error tolerance of 21 × 21 pixel and 20 × 20 pixel, respectively.Users can create their passwords according to the steps as shown in Fig. 3 (a).In addition, our system requires users to choose a location at zoom level of 16 due to the usability (similar to GeoPass).

User Study
To explore the performance of CPMap, we conducted another user study with a total of 50 participants, who did not attend the first study (in order to avoid any bias).Similarly, all participants are volunteers and have no any background in security.We gave an introduction about the tasks to each participant and asked them to sign a consent form before they started their work.The participants' background is detailed in Table 4.
In order to compare our scheme of CPMap with PassMap and GeoPass, we randomly divided the participants into two groups with 25 individuates each, named Group-A and Group-B.In particular, Group-A targets on a comparison between CPMap and PassMap, while Group-B focuses on CPMap and GeoPass.More implementation details of PassMap and GeoPass can refer to the former studies [30,32].To avoid any bias, we offered a guideline and trained all participants based on the same steps, ensuring that they understood the study steps and how to use these example systems.
Similar to our first study, each participant has 3 trials to get familiar with the assigned example systems.In the user study, every participant was asked to create 5 passwords for each scheme in their group.All participants should finish the experiments in the same day.In this case, a total of 250 trials were collected from each group during the whole study.The detailed steps in each experiment are summarized as follows: -Group-A.Participants in this group were required to create 5   User feedback.Regarding users' attitude, Table 6 summarizes the major questions and relevant scores (feedback) collected during this study.The first three questions attempt to investigate the creation experience regarding the different schemes, it is found that CPMap and GeoPass got a higher score than PassMap, i.e., 7.5 for PassMap but 8.8 for CPMap.The following three questions indicated that participants believed CPMap and GeoPass could provide better login experience (usability).For the last three questions, most participants believed PassMap and CPMap were more secure than GeoPass.In our informal interview, participants believed that two-step authentication could enhance the scheme security and increase the cracking difficulty for cyber-attackers, hence they considered CPMap to be more secure than GeoPass.Overall, most participants supported CPMap in terms of both security and usability.

Discussion and Limitations
-Security aspect.As mentioned earlier, CPMap is a two-step password scheme, which requires users to firstly choose a location on a world map and further click an object on an image.This aims to provide a better password space over GeoPass (with one clicked place on a map).In theory, PassMap should provide an even larger password space, but its security level would not be that high due to the weak password creation (refer to our first user study).
To provide a formal security analysis is one of our future work.-Usability aspect.Our study found that participants required much less time consumption in CPMap than that in PassMap.Actually, the time required by our scheme is quite close and even less than GeoPass (refer to Table 5).By analyzing the feedback collected from the participants, most participants considered both CPMap and GeoPass to be more usable PassMap.
As there are many GPs available in literature, one of our future work is to compare our scheme with other similar schemes like GeoPassNotes.

Conclusion
Map-based password authentication generally requires users to create their passwords by means of a (world) map.In this work, we firstly investigated how users would select two locations on a world map under the scheme of PassMap.It is found that common users may pick up two places that are very close to each other due to time considerations, which may greatly lower the security level.Motivated by the observation, we design CPMap, a click-points map-based GP scheme that allows users to choose one place on a world map at first and then click an object on an associated image relating to the selected location.We then conducted another user study with 50 participants to explore the scheme performance.Participants were found to perform better under our scheme as compared to PassMap and GeoPass in the aspects of both security and usability.Future work could include providing a thorough security analysis of password space and comparing our scheme with other similar schemes.

Fig. 1 .
Fig. 1.The implementation of PassMap: (a) registration page with user name, and (b) an example of selected locations.

Fig. 2 .
Fig. 2. The distribution of in-between distance for PassMap passwords in the study.

Fig. 3 .
Fig. 3. (a) The steps on how to create a CPMap password, and (b) an example: selection of one location on a world map in the first step and selection of one point on an image in the second step.

Table 1 .
Detailed information of participants in the user study.

Table 2 .
Table 2 computes the success rate and average completion time, which indicates that participants required around 36 seconds, 22 seconds, and 27 seconds Success rate and average completion time for the step of creation, confirmation and login in the study.

Table 3 .
Several main questions and relevant scores in the user study regarding PassMap.

Table 4 .
Detailed information of participants in the second user study.
passwords for each PassMap and CPMap, with a half hour rest in-between.The start from which scheme was selected by random.-Group-B.Participants in this group needed to create 5 passwords for GeoPass, and 5 passwords for CPMap after a half hour rest.The start from which scheme was selected by random.Participants from both groups should follow the same steps shown as below: • Step 1. Creation: creating a password following the related rules.• Step 2. Confirmation: confirming the password by inputting the same secrets in the correct place.If users incorrectly confirm their password, they can either retry this step or return to Step 1. • Step 3. Distributed memory: participants were provided with two finding tasks (paper-based) in order to distract them for 15 minutes.• Step 4. Login: logging into the example system with all created passwords.Users can cancel an attempt if they found an error.• Step 5. Feedback: participants are required to complete a feedback form about the scheme usage.Result analysis.Table 5 summarizes the success rate and average completion time regarding creation, confirmation and login for two groups.The main observations are discussed as below.-In Group-A, it is found that participants in CPMap could perform better than those in PassMap in the aspects of both login success and time consumption.For example, participants achieved a success rate of 78.4%, 80.8% and 82.4% for creation, confirmation and login in PassMap, but could increase the success rate to 86.4%, 88.0% and 89.6% in CPMap, respectively.Regarding time consumption, participants spent much less time in CPMap than those in PassMap, i.e., they spent 24.1 seconds for PassMap login, but only needed 11.6 seconds for CPMap login.-In Group-B, it is found that participants in CPMap could achieve a slightly better performance than those in GeoPass.For instance, participants reached a success rate of 84.0%, 87.2%, 89.6%, and 84.8%, 88.8%, 92.0% for GeoPass and CPMap, respectively.Regarding time consumption, these two schemes could achieve a similar result as well, i.e., they spent 14.4 seconds and 13.9 seconds for GeoPass and CPMap login.

Table 5 .
Success rate and average completion time for the step of creation, confirmation and login for two groups in the second study .

Table 6 .
Several main questions and relevant scores in the user study.