SOCIAL PSYCHOLOGY AND INFOSEC: Psycho-Social Factors in the Implementation of Information Security Policy M. E. Kabay, Ph.D. Director of Education, National Computer Security Association Carlisle, PA President, JINBU Corporation P. O. Box 509 / Westmount, QC H3Z 2T6 Canada Internet: email@example.com INTRODUCTION Security policies and procedures affect not only what people do but also how they see themselves, their colleagues and their world. Despite these psychosocial issues, security personnel pay little or no attention to what is known about social psychology. The established principles of human social behaviour have much to teach us in our attempts to improve corporate and institutional information security. Information security specialists concur that security depends on people more than on technology. Another commonplace is that employees are a far greater threat to information security than outsiders. It follows from these observations that improving security depends on changing beliefs, attitudes and behaviour, both of individuals and of groups. Social psychology can help us understand how best to work with human predilections and predispositions to achieve our goals of improving security: o research on social cognition looks at how people form impressions about reality (knowing these principles, we can better teach our colleagues and clients about effective security); o work on attitude formation and beliefs helps us present information effectively and so convince employees and others to cooperate in improving security; o scientists studying persuasion and attitude change have learned how best to change people's minds about unpopular views such as those of the security community; o studies of factors enhancing prosocial behaviour provide insights on how to foster an environment where corporate information is willingly protected; o knowledge of the phenomena underlying conformity, compliance and obedience can help us enhance security by encouraging compliance and by protecting staff against social pressure to breach security; o group psychology research provides warnings about group pathology and hints for working better with groups in establishing and maintaining information security in the face of ingrained resistance. The following discussion is based on well-established principles of social psychology. Any recent introductory college textbook in this field will provide references to the research that has led to the principles which are applied to security policy implementation. In this paper, references are to Lippa, R A (1990). Introduction to Social Psychology. Wadsworth (Belmont, CA). ISBN 0-534-11772-4. SOCIAL COGNITION Schemas are self-consistent views of reality. They help us pay attention to what we expect to be important and to ignore irrelevant data. They also help us organize our behaviour [Lippa, p. 141]. For example, our schema for relations at the office includes polite greetings, civil discussions, written communications, and businesslike clothes. The schema excludes obscene shrieks, abusive verbal attacks, spray-painted graffiti and colleagues dressed in swim suits. It is the schema that lets people tell what is inappropriate in a given situation. Security policies and procedures conflict with most people's schema. Office workers' schema includes sharing office supplies ('Lend me your stapler, please?'), trusting your team members to share information ('Take a look at these figures, Sally'), and letting your papers stay openly visible when you have to leave your desk. Unfortunately, sharing user IDs, showing sensitive information to someone who lacks the appropriate clearance, and leaving work stations logged on without protection are gross breaches of a different schema. Normal politeness dictates that when a colleague approaches the door we have just opened, we hold the door open for them; when we see a visitor, we smile politely (who knows, it may be a customer). In contrast, access policies require that we refuse to let even a well-liked colleague piggy-back their way through an access-card system; security policies insist that unbadged strangers be challenged or reported to security personnel. Common sense tells us that when the Chief Executive Officer of the company wants something, we do it; yet we try to train computer room operators to forbid entry to anyone without documented authorization--including the CEO. Schemas influence what we perceive [Lippa, p. 143]. For example, an employee refuses to take vacations, works late every night, is never late, and is never sick. A model employee? Perhaps, from one point of view. >From the security point of view, the employee's behaviour is suspect. There have been cases where such people have actually been embezzlers unable to leave their employment: even a day away might result in discovery. Saint or sinner? Our expectations determine what we see. Schemas influence what we remember [Lippa, p. 145]. When information inconsistent with our preconceptions is mixed with details that fit our existing schemas, we selectively retain what fits and discard what conflicts. When we have been fed a diet of movies and television shows illustrating the premise that information is most at risk from brilliant hackers, why should we remember the truth--that carelessness and incompetence by authorized users of information systems cause far more harm than evil intentions and outsiders ever do. Before attempting to implement policies and procedures, we should ensure that we build up a consistent view of information security among our colleagues. In light of the complexity of social cognition, our usual attempts to implement security policies and procedures seem pathetically inept. A couple of hours of lectures followed by a video, a yearly ritual of signing a security policy that seems to have been written by Martians--these are not methods that will improve security. These are merely lip service to the idea of security. According to research on counter-intuitive information, people's judgement is influenced by the manner in which information is presented. For example, even information contrary to established schemas can be assimilated if people have enough time to integrate the new knowledge into their world-views [Lippa, p. 148]. It follows that security policies should be introduced over a long time, not rushed into place. Preliminary information may influence people's responses to information presented later. For example, merely exposing experimental subjects to a list of words such as `reckless' or `adventurous' affects their judgement of risk-taking behaviour in a later test. It follows that when preparing to increase employee awareness of security issues, presenting case-studies is likely to have a beneficial effect on participants' readiness to examine security requirements. Pre-existing schemas can be challenged by several counter-examples, each of which challenges a component of the schema [Lippa, p. 153]. For example, prejudice about an ethnic group is more likely to be changed by contact with several people, each of whom contradicts a different aspect of the prejudiced schema. It follows that security awareness programs should include many realistic examples of security requirements and breaches. Students in the NCSA's Information Systems Security Course have commented on the unrealistic scenario in a training video they are shown; a series of disastrous security breaches occur in the same company. Based on the findings of cognitive social psychologists, the film would be more effective for training if the incidents were dramatized as occurring in different companies. Judgements are easily distorted by the tendency to rely on personal anecdotes, small samples, easily available information, and faulty interpretation of statistical information [Lippa, p. 155-163]. Basically, we humans are not rational processors of factual information. If security awareness programs rely strictly on presentation of factual information about risks and proposed policies and procedures, they will run up against our stubborn refusal to act logically. Security program implementation must engage more than the rational mind. We must appeal to our colleagues' imagination and emotion as well. We must inspire a commitment to security rather than merely describing it. Perceptions of risks and benefits are profoundly influenced by the wording in which situations and options are presented [Lippa, p. 163]. For example, experimental subjects responded far more positively to reports of a drug with `50% success' than to the same drug described as having `50% failure.' It follows that practitioners should choose their language carefully during security awareness campaigns. Instead of focusing on reducing failure rates (breaches of security), we should emphasize improvements of our success rate. BELIEFS AND ATTITUDES Psychologists distinguish between beliefs and attitudes. `A belief ... refers to cognitive information that need not have an emotional component....' An attitude refers to `an evaluation or emotional response....' [Lippa, p. 238]. Thus a person may believe that copying software without authorization is a felony while nonetheless having the attitude that it doesn't matter. Beliefs can change when contradictory information is presented, but some research suggests that it can take up to a week before significant shifts are measurable. Other studies suggest that when people hold contradictory beliefs, providing an opportunity to articulate and evaluate those beliefs may lead to changes that reduce inconsistency. These findings imply that a new concern for corporate security must be created by exploring the current structure of beliefs among employees and managers. Questionnaires, focus groups, and interviews may not only help the security practitioner, they may actually help move the corporate culture in the right direction. An attitude, in the classical definition, `is a learned evaluative response, directed at specific objects, which is relatively enduring and influences behaviour in a generally motivating way' [Lippa, p. 221]. The advertising industry spends over $50B yearly to influence public attitudes in the hope that these attitudes will lead to changes in spending habits--that is, in behaviour. Research on classical conditioning suggests that attitudes can be learned even because of simple word association [Lippa, p. 232]. If we wish to move our colleagues towards a more negative view of computer criminals, it is important not to portray computer crime using positive images and words. Movies like Sneakers may do harm indirectly by associating pleasant, likeable people with techniques that are used for industrial espionage. When teaching security courses, we should avoid praising the criminals we describe in case studies. One theory on how attitudes are learned suggests that rewards and punishments are important motivators. Studies show that even apparently minor encouragement can influence attitudes. A supervisor or instructor should praise any comments that are critical of computer crime or which support the established security policies. Employees who dismiss security concerns or flout the regulations should be challenged on their attitudes, not ignored. PERSUASION AND ATTITUDE CHANGE Persuasion--changing someone's attitudes--has been described in a terms of communications [Lippa, p. 258]. The four areas of research include o communicator variables: who is trying to persuade? o message variables: what is being presented? o channel variables: by what means is the attempt taking place? o audience variables: at whom is the persuasion aimed? Attractiveness, credibility and social status have strong effects immediately after the speaker or writer has communicated with the target audience; however, over a period of weeks to a month, the effects decline until the predominant issue is message content. We can use this phenomenon by identifying the senior executives most likely to succeed in setting a positive tone for subsequent security training. We should look for respected, likeable people who understand the issues and sincerely believe in the policies they are advocating. Fear can work to change attitudes only if judiciously applied. Excessive emphasis on the terrible results of poor security is likely to backfire, with participants in the awareness program rejecting the message altogether. Frightening consequences should be coupled immediately with effective and achievable security measures. Some studies suggest that presenting a balanced argument helps convince those who initially disagree with a proposal. Presenting objections to a proposal and offering counter-arguments is more effective than one-sided diatribes. The Software Publishers' Association training video, It's Just Not Worth the Risk, uses this technique: it shows several members of a company arguing over copyright infringement and fairly presents the arguments of software thieves before demolishing them. Modest repetition of a message can help generate a more positive response. Thus security awareness programs which include imaginative posters, mugs, special newsletters, audio and video tapes and lectures are more likely to build and sustain support for security than occasional intense sessions of indoctrination. The channel through which we communicate has a strong effect on attitudes and on the importance of superficial attributes of the communicator. `Face-to-face persuasion often proves to have more impact than persuasion through the mass media.... [because they] are more salient, personal and attention-grabbing, and thus they often stimulate more thought and commitment to their persuasive messages' [Lippa, p. 264]. Security training should include more than tapes and books; a charismatic teacher or leader can help generate enthusiasm for--or at least reduce resistance to--better security. Workers testing cognitive response theory [Lippa, p. 289] have studied many subtle aspects of persuasion. For example, experiments have shown that rhetorical questions (e.g., `Are we to accept invasions of our computer systems?') are effective when the arguments are solid but counter-productive when arguments are weak. In comparing the central route to persuasion (i.e., consideration of facts and logical arguments) with the peripheral (i.e., influences from logically unrelated factors such as physical attractiveness of a speaker), researchers find that the central route `leads to more lasting attitudes and attitude changes....' [Lippa, p. 293]. As mentioned above, questionnaires and interviews may help cement a favourable change in attitude by leading to commitment. Once employees have publicly avowed support for better security, some will begin to change their perception of themselves. As a teacher of information security, I find that I now feel much more strongly about computer crime and security than I did before I created my courses. We should encourage specific employees to take on public responsibility for information security within their work group. This role should periodically be rotated among the employees to give everyone the experience of public commitment to improved security. PROSOCIAL BEHAVIOUR Studies of how and why people help other people have lessons for us as we work to encourage everyone in our organizations to do the right thing. Why do some people intervene to stop crimes? Why do others ignore crimes or watch passively? Latane and Darley (Lippa, p. 493) have devised a schema that describes the steps leading to prosocial behaviour: o People have to notice the emergency or the crime before they can act. Thus security training has to include information on how to tell that someone may be engaging in computer crime. o The situation has to be defined as an emergency--something requiring action. Security training that provides facts about the effects of computer crime on society and solid information about the need for security within the organization can help employees recognize security violations as emergencies. o We must take responsibility for acting. The bystander effect comes into play at this stage. The larger the number of people in a group confronted with an emergency, the slower the average response time. Larger groups seem to lead `to a diffusion of responsibility whereby each person felt less personally responsible for dealing with the emergency' [Lippa, p. 497]. Another possible factor is uncertainty about the social climate; people fear `appearing foolish or overly emotional in the eyes of those present.' We can address this component of the process by providing a corporate culture which rewards responsible behaviour such as reporting security violations. o Having taken responsibility for solving a problem, we must decide on action. Clearly written security policies and procedures will make it more likely that employees act to improve security. In contrast, contradictory policies, poorly-documented procedures, and inconsistent support from management will interfere with the decision to act. Another analysis proposes that people implicitly analyze costs of helping and of not helping when deciding whether to act prosocially. The combination of factors most conducive to prosociality is low cost for helping and high cost for not helping. Security procedures should make it easy to act in accordance with security policy; e.g., there should be a hot-line for reporting security violations, anonymity should be respected if desired, and psychological counselling and followup should be available if people feel upset about their involvement. Conversely, failing to act responsibly should be a serious matter; personnel policies should document clear and meaningful sanctions for failing to act when a security violation is observed; e.g., inclusion of critical remarks in employment reviews and even dismissal. One method that does not work to increase prosocial behaviour is exhortation [Lippa, p. 513]. That is, merely lecturing people has little or no effect. On the other hand, the general level of stress and pressure to focus on narrow tasks can significantly reduce the likelihood that people will act on their moral and ethical principles. Security is likely to flourish in an environment that provides sufficient time and support for employees to work professionally; offices where everyone responds to self-defined emergencies all the time will not likely pay attention to security violations. Some findings from research confirm common sense. For example, guilt motivates people to act more prosocially. This effect works best `when people are forced to assume responsibility....' Thus enforcing standards of security using reprimands and sanctions can indeed increase the likelihood that employees will subsequently act more cooperatively. In addition, mood affects susceptibility to prosocial pressures: bad moods make prosocial behaviour less likely, whereas good moods increase prosociality. A working environment in which employees are respected is more conducive to good security than one which devalues and abuses them. Even cursory acquaintance with other people makes it more likely that we will help them; it thus makes sense for security supervisors to get to know the staff from whom they need support. Encouraging social activities in an office (lunch groups, occasional parties, charitable projects) enhances interpersonal relationships and can improve the climate for effective security training. CONFORMITY, COMPLIANCE AND OBEDIENCE Turning a group into a community provides a framework in which social pressures can operate to improve our organization's information security. People respond to the opinions of others by (sometimes unconsciously) shifting their opinion towards the mode. Security programs must aim to shift the normative values (the sense of what one should do) towards confidentiality, integrity and availability of data. As we have seen in public campaigns aimed at reducing drunk driving, it is possible to shift the mode. Twenty years ago, many people believed that driving while intoxicated was amusing; today a drunk driver is a social pariah. We must move towards making computer crime as distasteful as public drunkenness. The trend towards conformity increases when people within the group like or admire each other [Lippa, p. 534]. In addition, the social status of an individual within a group influences that individual's willingness to conform. High-status people (those liked by most people in the group) and low-status people (those disliked by the group) both tend to more autonomous and less compliant than people liked by some and disliked by others [Lippa, p. 536]. Therefore the security officers should pay special attention to those outliers during instruction programs. Managers should monitor compliance more closely in both ends of the popularity range. Contrariwise, if security practises are currently poor and we want allies in changing the norm, we should work with the outliers to resist the herd's anti-security bias. 'The norm of reciprocity holds that we should return favours in social relations' [Lippa, p. 546]. Even a small, unexpected or unsolicited (and even unwanted) present increases the likelihood that we will respond to requests. A security awareness program that includes small gifts such as a mug labelled `SECURITY IS EVERYONE'S BUSINESS' or an inexpensive booklet such as the Information Systems Security Pocket Guide (available from the NCSA) can help get people involved in security. The `foot in the door' technique suggests that we `follow a small initial request with a much larger second request' [Lippa, p. 549]. For example, we can personally ask an employee to set a good example by blanking their screen and locking their terminal when they leave their desk. Later, once they have begun their process of redefinition of themselves ('I am a person who cares about computer security'), we can ask them for something more intense, such as participating in security training for others (e.g., asking each colleague to blank their screen and lock their terminal). GROUP BEHAVIOUR Early studies on the effects of being in groups produced contradictory behaviour; sometimes people did better at their tasks when there were other people around and sometimes they did worse. Eventually, social psychologist Robert Zajonc [Lippa, p. 572 ff.] realized that `The presence of others is arousing, and this arousal facilitates dominant, well-learned habits but inhibits nondominant, poorly-learned habits.' Thus when trying to teach employees new habits, it is counter-productive to put them into large groups. Individualized learning (e.g., computer-based training, video tapes) can overcome the inhibitory effect of groups in the early stages of behavioural change. Another branch of research in group psychology deals with group polarization. Groups tend to take more extreme decisions than individuals in the group would have [Lippa, p. 584]. In group discussions of the need for security, polarization can involve deciding to take more risks--by reducing or ignoring security concerns--than any individual would have judged reasonable. Again, one-on-one discussions of the need for security may be a more effective approach to building a consensus that supports cost-effective security provisions than large meetings. In the extreme, a group can display groupthink, in which a consensus is reached because of strong desires for social cohesion [Lippa, p. 586 ff.]. When groupthink prevails, evidence contrary to the received view is discounted; opposition is viewed as disloyal; dissenters are discredited. Especially worrisome for security professionals, people in the grip of groupthink tend to ignore risks and contingencies. To prevent such aberrations, the leader must remain impartial and encourage open debate. Experts from the outside (e.g., respected security consultants) should be invited to address the group, bringing their own experience to bear on the group's requirements. After a consensus has been achieved, the group should meet again and focus on playing devil's advocate to try to come up with additional challenges and alternatives. CONCLUSIONS By viewing information security as primarily a management issue, we can benefit from the mass of knowledge accumulated by social psychologists. We can implement security policies and procedures more easily by adapting our training and awareness techniques to correspond to human patterns of learning and compliance. SUMMARY OF RECOMMENDATIONS 1. Before attempting to implement policies and procedures, we should ensure that we build up a consistent view of information security among our colleagues. 2. Security policies should be introduced over a long time, not rushed into place. 3. Presenting case-studies is likely to have a beneficial effect on participants' readiness to examine security requirements. 4. Security awareness programs should include many realistic examples of security requirements and breaches. 5. We must inspire a commitment to security rather than merely describing it. 6. Emphasize improvements rather than reduction of failure. 7. A new concern for corporate security must be created by exploring the current structure of beliefs among employees and managers. 8. Do not to portray computer crime using positive images and words. 9. Praise any comments that are critical of computer crime or which support the established security policies. 10. Employees who dismiss security concerns or flout the regulations should be challenged on their attitudes, not ignored. 11. Identify the senior executives most likely to succeed in setting a positive tone for subsequent security training. 12. Frightening consequences should be coupled immediately with effective and achievable security measures. 13. Presenting objections to a proposal and offering counter-arguments is more effective than one-sided diatribes. 14. Security awareness programs should include repeated novel reminders of security issues. 15. In addition to tapes and books, rely on a charismatic teacher or leader to help generate enthusiasm for better security. 16. Encourage specific employees to take on public responsibility for information security within their work group. 17. Rotate the security role periodically. 18. Security training should include information on how to tell that someone may be engaging in computer crime. 19. Build a corporate culture which rewards responsible behaviour such as reporting security violations. 20. Develop clearly written security policies and procedures. 21. Security procedures should make it easy to act in accordance with security policy. 22. Failing to act in accordance with security policies and procedures should be a serious matter. 23. Enforcing standards of security can increase the likelihood that employees will subsequently act more cooperatively. 24. A working environment in which employees are respected is more conducive to good security than one which devalues and abuses them. 25. Security supervisors should get to know the staff from whom they need support. 26. Encourage social activities in the office. 27. Pay special attention to social outliers during instruction programs. 28. Monitor compliance more closely in both ends of the popularity range. 29. Work with the outliers to resist the herd's anti-security bias. 30. Include small gifts in your security awareness program. 31. Start improving security a little at a time and work up to more intrusive procedures. 32. Before discussing security at a meeting, have one-on-one discussions with the participants. 33. Remain impartial and encourage open debate in security meetings. 34. Bring in experts from the outside when faced with groupthink. 35. Meet again after a consensus has been build and play devil's advocate.
Please report problems with the web pages to the maintainer