A framework for evaluating security risks associated with technologies used at home.
Computation is embedded throughout our homes. Some devices are obvious: desktops, laptops, wireless routers, televisions, and gaming consoles. Increasingly, however, computational capabilities are appearing in our appliances, healthcare devices, children's toys, and the home's infrastructure. These devices are incorporating new sensors, actuators, and network capabilities: a Barbie with a video camera1; a lock for your front door controlled by your cell phone; or a bathroom scale that reports readings over your wireless network.26 Many of these devices are also subject to control by servers external to the home, or are mobile technologies that regularly leave the home's perimeter and interact with other networks. These trends, which we expect to accelerate in the coming years, create emergent threats to people's possessions, well-being, and privacy. We seek to survey the security and privacy landscape for devices in the home and provide a strategy for reasoning about their relative computer security needs.
Many human assetswhether electronic, physical, or nontangible items of value to end userscan be accessed or influenced from computing devices within the home; unsurprisingly, these assets are also potentially attractive targets to adversaries. The capabilities of new electronics and their presence in the home facilitate traditional crimes and allow new classes of attacks. Technically savvy burglars, for example, may use technology both to identify houses with expensive, easily resold items and to better plan and execute their crimes. Adversaries can also target technologies with a wide range of new capabilities, with the goal of accessing video and audio feeds,25 unlocking doors or disabling home security,27 tampering with home healthcare devices,13,26 or interfering with home appliances and utilities.22,24
Fortunately, there have been few "high tech" crimes to date exploiting these new capabilities. Now is the right time to develop a foundation for securing the myriad devices within the home: before these technologies become more ubiquitous, communicative, and capable, and before real adversarial pressures emerge. While progress has been made in understanding security concerns for specific home technologies or categories of technologies,3,14 there is currently a lack of unified vision for evaluating security threats posed by the assortment of consumer devices within the home. There are trade-offs in the design of any security system, but without a cohesive strategy for reasoning about home device security, product manufacturers will be left to determine the appropriate trade-offs for themselves without best-practice references.
Our goal is first to survey the landscape of potential attacks, then to provide structure and guidance for reasoning about the differing security needs of home technologies. While many of these elements will be familiar from security for traditional computers, their implications are worth reassessing in the context of the home ecosystem. This article is also complementary to an existing body of research on security for home technologies, including work on the security needs and behaviors of users4,8,10,18,19 and work on centralized security technology solutions.28
Table 1 presents an overview of the topics covered by this article. We begin by presenting an overview of how the ecosystem of home technologies can enable a range of attacks with electronic and physical consequences. Building on this discussion, we present the two key components of our strategy for evaluating the potential risks with home technologies: a taxonomy of security goals for home technologies; and a set of device characteristics that can be used to estimate a device's potential risk to users. We apply our approach to three home technologies: a webcam toy, a networked scale, and a home automation device. Our framework is not intended to be definitive, but rather informative: our intent is that this approach will provide a useful starting point for home technology stakeholders ranging from product manufacturers to consumer advocacy groups to the research community. Moreover, by focusing on the entire home technology ecosystem, our hope is that this work will strengthen the foundations for developing secure home technologieswith the ultimate goal of creating a trustworthy home environment for users.
The home technology space is interesting and unique from other domains. In a nutshell, the new home landscape takes four challengeschallenges that are not unique in and of themselvesand combines them to create a new problem space: (1) an extremely personal, asset-filled environment where there is (2) no dedicated, professional administrator to maintain a (3) heterogeneous collection of consumer technologies that (4) are increasingly cyber-physical and sensor-rich. The combination of these factors leads to an array of attacks and complicates the design of defenses for home devices. From a technical perspective, the home is filled with a diverse range of technologies with varying levels of security, hybrid communication structures, and no centralized security management system. From a human perspective, the home contains private and semi-private spaces shared by children, parents, siblings, elderly, roommates, and guests. Interpersonal dynamics, varying levels of security expertise, and different social and technical preferences all contribute to complicating the home technology security landscape. In order to effectively create and evaluate defenses, it is important to first understand the threat landscape.
Attack Scenarios. One unique aspect of the new home technology space is the vast array of attacks that it enablesmany of which differ in effect from Web or desktop attacks. The increasing presence of electronics in the homecontrolling our houses and coordinating our livesprovides unique opportunities for the technically savvy criminal.
Table 2 breaks down attacks into three tiers: low-level mechanisms, intermediate goals, and high-level goals. The low-level mechanisms listed in Table 1such as denial-of-service attacks, tampering with logs, or eavesdropping on network trafficwill be familiar to anyone who has experience with computer security. However, the additional focus on sensors and actuators is something that is not generally encountered with traditional computing devices. Similarly, the high-level goals behind the attacks (blackmail, extortion, theft, and vandalism, among others) are the same motivations that one encounters with all criminal activities. Arguably, the most novel aspects of attacks on the home ecosystem are the intermediate goals: the ways in which the unique capabilities of devices or the assets to which they have access enable criminal opportunities.
In order to highlight some of the unique properties of the home ecosystem, we list examples of attacks that are not viable with traditional computing platforms:
Attack targets. For many types of attacks, an adversary could either attempt to target a particular person of interest or simply take advantage of known hardware and software flaws to indiscriminately attack any vulnerable victim. Attacks on a designated person require that the adversary identify useful exploits for the target's particular technology configuration. On the other hand, for attacks on "low-hanging" targetsattacks of exploitative opportunitythe adversary need only focus on a known exploit and locate victims who are vulnerable to that exploit.
The physical and the electronic. At a high level, it is interesting that the presence of actuators and sensors in the new home environment allows interactions between the physical and electronic states of devices. It is possible to perform electronic attacks with physical consequences, but it is also possible to perform physical attacks with electronic consequences, or attacks that have both physical and electronic components. As an example of a physical attack that has electronic (then physical) consequences, an adversary might apply a bright, directed light source to an external light sensor in order to trick outdoor flood lighting into turning off. Similarly, one can imagine an attack where physically tricking a system sensor causes the system to enter a fail-safe mode that is more easily compromised via electronic attack.
Infection Pathways. The challenges of the home environmentsuch as its heterogeneous topology and the idiosyncrasies of its occupantshelp enable novel or complex infection pathways. Mobile devices, infrastructure electronics, cyber-physical systems, guest devices, and machines brought home from work all commingle in one hodgepodge environment, increasing the exposure to compromise. Understanding the potential infection pathwaysparticularly nontraditional pathwaysthat malware might follow to compromise a device helps us understand its exposure to risk, which we use later in our characterization of device risk. The Infection Pathways column of Table 1 provides an overview of the kinds of pathways that malware can take to infect a device in the home.
Entry points. There are a number of entry points an adversary could use to attack home technologies. Electronically, a device on the home network might be compromised by a direct attack from a device external to the home, or compromised by an infected device within the home (whether stationary, mobile, or belonging to a guest). If a device is mobile and connects to an infected network, it might become infected. Physically, a device might be infected by a manual interface such as USB or CD.5,9 Alternative physical attack vectors include: receiving an infected device as a gift; purchasing a used, compromised device from a source such as eBay or Craigslist; purchasing a "new" device that has previously been purchased, infected, then returned; or purchasing a device that was infected during its manufacture.11 Additionally, an adversary has a number of opportunities to socially engineer a user into installing malware, such as via app stores.15,21 As another vector, an adversary could take advantage of the increasing number of "prosumers"consumers who jailbreak their devices or perform similar automated modificationswhose devices allow behaviors that go beyond the capabilities expected by the manufacturer's typical APIs and might not receive security software updates.
Stepping back. As this survey of the attack scenarios and infection pathways shows, the risks with computer security vulnerabilities in home technologies are quite varied and, in some cases, significant. Here, we present a framework for more methodically identifying and prioritizing the security risks within the home.
To design a system for defending home technologies, it is necessary to understand the human assets that are at stake and the desired security goals. We present a casual taxonomy of goals for protecting human assets in the home (also shown in the Defensive Goals column in Table 1). The general goals of confidentiality, integrity, authenticity, and availability are familiar security concepts; we frame the goals for defending the home slightly differently in order to highlight the domain in which they are applied and the unusual consequences of security failures. This taxonomy is meant to approach security and privacy goals from a variety of perspectives, and as such items are not mutually exclusive.
Security failures can result in a variety of kinds of harm to users. It is common to consider harm to users in terms of financial assets; it is less typical to consider damaging users by, for example, wasting their time or causing them stress. We suggest considering the potential negative impact of attacks on the following assets (in the Human Assets column in Table 1): emotional well-being, financial well-being, personal data, physical well-being, and relationships. In addition to considering the assets of individuals, it can be beneficial to consider the broader assets of societal well-being and impact on the biosphere. The list is derived in part from Value Sensitive Design12an area of human-computer interaction that focuses on what different individuals valueand in part from the discussion sections of papers on emerging technologies.5,7,16
Device Goals. These are security goals that pertain to the operation of traditional or embedded computing devices.
Digital Data Goals. These are security goals that pertain to a user's digital data.
Environment Goals. We must also consider security goals that pertain to the home infrastructure and general environmental conditions.
Having explored human assets and security goals, we now explore a strategy for evaluating the potential risks with home technologies.
The risk posed by a given home technology can be broken down into three components: the feasibility of an attack on the system; the attractiveness of the system as a compromised platform; and the damage caused by executing a successful attack. The first two factors, when combined, provide some indication of the likelihood that an adversary will compromise the device in question, while the third factor helps weight the overall risk. The human assets and security goals discussed previously provide a framework for reasoning about the impacts of potentially successful attacks. Here, we provide some guidelines for how to evaluate a device's exposure to attack and the likelihood of an attack attempt based upon the rough design characteristics of a technology (also summarized in the Device Risk Axes column of Table 1). Such a strategy for evaluation could be used by product designers, policymakers, or consumer advocacy groups.
Potential Exposure to Attack. In order to determine the risk posed by a home technology, it is necessary to evaluate how vulnerable the device is to an attack. It is difficult to make arbitrary evaluations of a technology's vulnerability without performing a hands-on study of the device in question; nonetheless, we provide some loose guidelines for design factors that tend to increase the likelihood that a device may be vulnerable to compromise. Those devices that are most likely to be vulnerable may deserve the most security consideration.
If a device is mobile, then the chances are higher that it will come into contact with malicious or infected networks or devices.
We stress that these guidelines indicate the likelihood of a potential vulnerability absent appropriate defenses, and are not an absolute measure of risk. Second, we stress that the list here is not exhaustive: instead, it focuses on some common issues that affect a device's attack surface. One would need to conduct a full security analysis of a product in order to more accurately gauge its level of security.
Communication capabilities. The more communication capabilities that a device possesses (for example, Wi-Fi, Ethernet, infrared, Bluetooth, ZigBee, cellular, powerline), the more media an adversary can use to attack the device. Manual communications capabilities such as USB or CD interfaces must also be considered.
Communication behavior. We consider three aspects of a device's communication behavior that affect its exposure to attack: initiated communications; receptiveness to incoming communications; and mobility. If a device is designed to communicate with a server or peer external to the home network, then a remotely located adversary has increased opportunities to attempt a range of passive and active attacks such as traffic eavesdropping, man-in-the-middle attacks, relay attacks, replay attacks, and spoofing. Additionally, a device's receptiveness to acting upon or replying to incoming network communications may also increase its exposure to attack.
If a device is mobile, then the chances are higher that it will come into contact with malicious or infected networks or devices.
The cloud. The flexibility and affordability of storage and computation in the cloud (such as software-as-a-service, platform-as-a-service, infrastructure-as-a-service) are causing more manufacturers to rely on the cloud for storage, back-up, remote access, or configuration. If data is stored on those remote servers, then we must consider the risks to users if that data is exposed, altered, rendered inaccessible, or otherwise misused. By facilitating online configuration or remote access, manufacturers expose a different surface to attackone that should not be overlooked even though it lies outside the physical boundaries of the home.
Software updates. The ability or inability to perform software updates can have positive or negative implications in a security context.2 A device that connects to a manufacturer's server regularly to download updates may receive patches that remove vulnerabilities; however, if the update system does not properly verify that an update is legitimate or if that verification process is flawed, then an adversary has a convenient mechanism with which to modify a device's behavior.
Configuration defaults, user interfaces, and users. Defaults, user interfaces, and intended users all affect a device's security configuration (for example, sharing settings, account passwords, or update settings) and therefore its ultimate vulnerability to attack. A device with more secure default settings has an advantage over devices with less secure defaults, as some users never modify default configurations. Entire research venues are dedicated to tackling issues surrounding configuration models and defaults.
Some user interfaces are rich whereas others are minimal. There are advantages and disadvantages with each. Rich interfaces have the potential to be confusing but can allow greater control over security settings. Rich interfaces can also inform users of security compromises and give them the ability to respond.
Similarly, it is important to consider the characteristics of the people who are most likely to administer the device. Different users might have different levels of security caution, different levels of familiarity with computers, or different priorities. For example, if a device resembles a toy or is meant to be used by children, then parents might give it to their children to administer, despite the child's likely lack of experience with computer security and different stance on privacy issues.
Attractiveness as a Target. To understand the risk posed by a home technology, it is also necessary to consider how much value the device holds for an adversary. A device's attractiveness to an adversary is relevant for two reasons: first, it affects the likelihood that an adversary will attempt to compromise the device. Second, the properties that cause a device to be of interest to an adversary are most likely the same properties that make the device a potential risk to users: after all, an adversary has some goal in attacking the device, and most of those goals cause direct or indirect harm to the user. We articulate here some of the capabilities and usage scenarios that make a device more attractive as an attack target.
Technology market share. If an adversary is intending to perform attacks of exploitative opportunityattacks targeted at nonspecific vulnerable people rather than specific victimsthen it is most efficient for the adversary to attack a technology that is deployed in many homes. Conversely, targeted attacks may better succeed with devices that have received less scrutiny due to a smaller market share.
Intended users and usage. Understanding a technology's most likely usage scenario helps indicate how valuable it would be to an adversary, since it dictates the assets with which the technology will interact. For example, a nanny cam would allow an adversary to spy on children; a networked storage server might hold backups of tax records or other financial data; and an electronically controlled door lock might allow full access to a home, whereas an electronic garage door opener would only allow access to the garage. While one cannot always anticipate how a device might be repurposed, it is important to consider future usage scenarios.
Sensors. If a device has sensors that record data then it might be a target of increased interest. The value of a sensor depends upon how much interest the raw or mined data holds for the adversary: for example, microphones and cameras have obvious value for voyeurs, blackmailers, or even private investigators or industrial spies; accelerometers might indicate whether or not a person is awake; and devices with GPS or Wi-Fi can be used to track an individual.
Actuators. A device holds increased value for an adversary if it can be used to effect changes in the physical world, since cyber-physical systems are both more efficient and less risky to use than physically traveling to a home. Cyber-physical effects of interest might include: locking or unlocking doors, cutting off electricity or water, changing thermostat temperatures, controlling lights, and turning appliances such as fireplaces on or off.
Power. The power reserves and power schedule of a device affects its utility to an adversary. A device with limited battery life, such as a mobile phone or a universal Wi-Fi remote, has constraints on its usefulness. Alternatively, devices that are regularly unplugged or powered off by their users are not dependably accessible to the adversary.
Connectedness. A target might have value for an adversary either because it is likely to interact with many devices in the futuredue to mobility or high network trafficor because it will interact with a device of particular interest to the attacker; for example, an adversary might target a mobile device with the intention that it will later be able to infect networked-attached storage that houses financial data.
Storage and computation. Devices with large storage capabilities might be targeted to store illegal materials. Devices with smaller storage capabilities are less useful on their own, but could be used as part of a distributed storage botnet. Devices with large computational capabilities might also be attractive to adversaries with heavy computational tasks, such as farming Bitcoins or cracking passwords. While there are additional properties that might affect a device's potential exposure to attack or its attractiveness as a target for attack, we chose to list the characteristics that we judged most significant and relevant for home technologies.
We tie together our framework with an example of how one might use it to analyze or compare the potential risks posed by different technology designs. We present a conceptual investigation of three technologies: a mobile webcam toy, a wireless scale, and a siren for a home security system. These technologies are not meant to be specific products, but rather amalgamations of products or exemplars of product categories. They represent a range of target audiences, technical capabilities, and application scenarios.
Tables 3 and 4 present a high-level view of how our framework might be used to evaluate the approximate risk posed by these device designs. Interpretations and rankings of different risk levels are subjective and depend upon perspective. Table 3 considers the technologies according to the characteristics presented in the section "Evaluating Potential Risks." Table 4 summarizes the consequences that can result if the security goals discussed in the section "Human Assets and Security Goals" are not met. Color-coding provides an overview of the comparative risk patterns of the different devices.
Our goal is not to be exhaustive or predictive. Rather, our goal is to facilitate an informed discussion about the potential risks with a technology if security is not sufficiently addressed in its design. To clarify, this framework only provides a skeleton for characterizing risks; individuals not accustomed to considering attack scenarios might require additional guidance.
Mobile webcam. Having populated the tables, we can now quickly assess the potential security risks with each technology. With its communications capabilities, communication behaviors, and user interface design, the mobile webcam toy clearly has significant potential exposure to attack (Table 3). Furthermore, with its proximity to children and its significant sensing capabilities (camera and microphone), the webcam toy appears to be a potentially attractive target to some adversaries (Table 3); more particularly, this device might be an attractive target to adversaries seeking to compromise the privacy of home occupants (Table 4). Given the high potential exposure to compromise, the attractiveness of the target, and the importance of the corresponding security goals, we would identify the mobile toy robot as a technology that merits significant security review by product designers before the device enters the market. Similarly, based on the data in these tables, consumer advocacy groups would likely identify this device as one deserving post-market security auditing.
Fortunately, security best practicesif deployedcould significantly harden this device against attack: for example, the ability to perform authenticated software updates could allow the manufacturer to quickly address vulnerabilities once uncovered and strong audit logs could help further dissuade attack.
Wireless scale. Turning to the wireless scale, we see that although it does have some technical features that increase its potential exposure to attacks (Table 3)particularly the inclusion of Wi-Fi capabilitiesit is not a particularly attractive attack target and the associated security goals are not critical (Table 4). While there are arguments for trying to harden all devices against all possible attacks, that strategy is not feasible in practice. First, increasing security may impact the usability, desirability, or utility of the product. Second, companies do not have unlimited budgets to spend on security. These tables suggest that if a single manufacturer produced both the mobile webcam toy and the scale, the company would be well advised to focus its security efforts on the webcam toy over the wireless scale; nevertheless, given the scale's potential effects on emotional well-being, eating, or exercise activities, the integrity of sensor readings might become a security priority if the product were being marketed toward users with eating disorders (Table 4).
Security siren. Finally, we turn to the security siren. Table 4 suggests the primary security goals for the siren are related to device operability and command authenticity. If an attacker can disable the siren, then the attacker might be able to enter a home without alerting those nearby, thereby rendering the short-term benefits of the home alarm system ineffective; the home security system might still automatically call the police, but the police will not arrive immediately. Since the market share is listed as small in Table 3, the likelihood of an attacker choosing to target this system today seems small; however, the market share may increase over time. Having identified device operability as a particularly pertinent security goal, the device manufacturer can once again implement techniques to harden the device. For example, the device could issue a distinctive alert if a denial-of-service attack renders the siren unavailable to the rest of the home automation network. The continuous sounding of an alarm could also cause a service interruption by tempting the user into turning off or ignoring the system; therefore, it is also important for the manufacturer to deploy defenses such as transmitting logs and incident reports to a monitoring agency.
Stepping back. As these examples illustrate, our framework can guide the analysis of potential security risks with technologies in the home. Devices in the home will likely incorporate varying degrees of security defenses, due in part to oversights by designers and developers, but also due to the costs associated with implementing security measures. By methodically evaluating a device's potential exposure to attack and its attractiveness to adversaries (Table 3), as well as the potential impacts on security goals and human assets if the device is compromised (Table 4), one can assess the degree to which security might be important for a given device, as well as which security goals are the most important to address. This information can help developers focus their energies on the most significant risks of a design and help consumer advocacy groups direct their attention toward the computer security properties of the most concerning home technologies.
Our homes are increasingly becoming hubs for technologies with a wide variety of capabilities. While it would be ideal to strive for "perfect" security on all consumer devices, the reality is that resources such as time and money constrain these efforts. In the coming years, it will become increasingly important to improve the efficacy, interoperability, and usability of computer security solutions for the home. It remains to be seen what such a security solution would look like. It might take the form of a centralized security console that displays and controls device permissions and traffic.28 The security system could incorporate trusted hardware, network intrusion detection systems, tiered security,6,20 or cryptographic trust evidence of past transactions or device state.
We need a strategy for how to secure devices in the home. We need to understand the potential risks: risks that are a function of a device's potential exposure to attack, its attractiveness as an attack target, and the potential impacts on human assets if the device is compromised. In this article, we explored the landscape of technological attacks on the home and provided a strategy for thinking about security in the home. In particular, we have identified human assets at stake within the home and security goals for computational home devices. We then identified key features of devices that, in general, make them more vulnerable to attack or more attractive as attack targets. Together, these axes can be used to evaluate the level and type of security attention appropriate for different home technologies. We applied our approach to three example technologies: a wireless webcam toy, a wireless scale, and a home automation siren. With further research, we conjecture that our risk framework could be distilled into a decision tree-like structure with questions that would allow those without security expertise to deterministically assign a device to a risk category. By seeking to understand the risks posed by home technologies as a cohesive whole, our hope is that this work will strengthen the foundations for developing secure home technologieswith the ultimate goal of creating a more trustworthy home environment for users.
We thank Intel and the Intel Trust Evidence Program for supporting this work. We thank Dan Halperin, Greg Piper, Jesse Walker, and Meiyuan Zhao for feedback on earlier versions of this article.
1. Barbie Video Girl; http://www.barbie.com/videogirl/
5. Checkoway, S., McCoy, D., Kantor, B., Anderson, D., Shacham, H., Savage, S., Koscher, K., Czeskis, A., Roesner, F. and Kohno, T. Comprehensive experimental analyses of automotive attack surfaces. In Proceedings of USENIX Security '11.
6. Cisco NAC; http://www.cisco.com/en/US/products/ps6128/index.html.
9. Edwards, C., Kharif, O. and Rile, M. Human Errors Fuel Hacking as Test Shows Nothing Stops Idiocy (June 27, 2011); http://www.bloomberg.com/news/2011-06-27/human-errors-fuel-hacking-as-test-shows-nothing-prevents-idiocy.html
11. Fisher, D. Samsung Handsets Distributed With Malware-Infected Memory Cards (June 4, 2010); http://threatpost.com/en_us/blogs/samsung-handsets-distributed-malware-infected-memory-cards-060410
13. GlowCaps; http://www.rxvitality.com/glowcaps.html.
15. Greenberg, A. iPhone Security Bug Lets Innocent-Looking Apps Go Bad (Nov. 7, 2011); http://www.forbes.com/sites/andygreenberg/2011/11/07/iphone-security-bug-lets-innocent-looking-apps-go-bad/
16. Halperin, D., Heydt-Benjamin, T.S., Ransford, B., Clark, S.S., Defend, B., Morgan, W., Fu, K., Kohno, T. and Maisel, W.H. Pacemakers and implantable cardiac defibrillators: Software radio attacks and zero-power defenses. IEEE 2008.
19. Mazurek, M.L., Arsenault, J.P., Bresee, J., Gupta, N., Ion, I., Johns, C., Lee, D., Liang, Y., Olsen, J., Salmon, B., Shay, R., Vaniea, K., Bauer, L., Cranor, L.F., Ganger, G.R. and Reiter, M.K. Access control for home data sharing: Attitudes, needs and practices. In Proceedings of CHI '10.
20. Microsoft NAP; http://technet.microsoft.com/en-us/network/bb545879.
21. Mills, E. More malware targeting Android (July 11, 2011); http://news.cnet.com/8301-27080_320078606-245/more-malware-targeting-android/
22. Nest; http://www.nest.com/.
23. Patel, S.N., Reynolds, M.S. and Abowd, G.D. Detecting human movement by differential air pressure sensing in HVAC system ductwork: An exploration in infrastructure mediated sensing. In Proceedings of Pervasive '08.
24. Rock Star in your kitchen; (Aug. 29, 2008); http://www.gorenjegroup.com/en/news?aid=933
25. Spykee; http://www.spykeeworld.com/.
26. Withings WiFi Body Scale; http://www.withings.com/en/bodyscale.
27. XFINITY Home Security; http://www.comcast.com/homesecurity/.
Table 3. An approximate risk evaluation of the three example technologies via potential exposure to attack and attractiveness of the attack target. The cells are color-coded to indicate the approximate severity of the concern: dark orange (serious), light orange (moderate), and light blue (minor).
Table 4. An approximate risk evaluation of the three example technologies considering how human assets might be impacted if defensive goals are not met. The cells are color-coded to indicate the approximate severity of the concern: dark orange (serious), light orange (moderate), and light blue (minor).
©2013 ACM 0001-0782/13/01
Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and full citation on the first page. Copyright for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers, or to redistribute to lists, requires prior specific permission and/or fee. Request permission to publish from email@example.com or fax (212) 869-0481.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2013 ACM, Inc.
No entries found