Members

Matthew Chou
Mike Preston
Thomas McMahon
David Barrera

Note: research so far moved to Discussion section.

Abstract

The goal of this article is to investigate the feasibility of implementing a system of justice on a distributed computing environment. Although directly applying human concepts related to justice, for example intent, is not possible in the realm of computers, we can use these concepts to construct a justice system that helps maintain the stability and efficiency of the distributed environment. To provide this functionality, the justice system requires a reporting mechanism as well as a mechanism for guaranteed attribution of transactions such that members of the distributed society can flag deviant behaviour and proper punishment may be exacted based on collected evidence.

This article is divided into two main sections; the first is a discussion on theories of human justice and how concepts from the human example may be used within the scope of computers. The second section describes the components necessary to create a justice system for a distributed computing society and how these components would be used in four deviant acts; comment spam, unauthorized access, denial of service, and phishing attacks.

Can Justice be Implemented on a Distributed Computing System: Discussion

In this section we present definitions of human justice and punishment. It is important to understand these concepts so that we can use them as a template to create a system of justice for a distributed computing environment. This section also includes definitions of key concepts related to justice and how they relate to potential justice in the realm of computers.

What is Justice?

Theory of Justice

John Rawls provides a definition of the purpose of justice as providing two primary functions; first, justice assigns rights and duties for the basic institutions of society, and second, justice describes the best way to distribute the benefits and burdens of society. [1] Essentially justice must ensure that a society is able to operate efficiently and with sufficient stability. This view fits well within the scope of justice for a distributed computing system as there are very clear roles which can be assigned and there are finite resources which can be used to manage the “benefits and burdens” of society.

In order for a society to uphold justice it must possess the ability to punish those who behave in a deviant manner. From a philosophical point of view, there are three main categories of punishment; Teleologic, Retributive, and Teleologic Retributive. [3]

Teleologic View of Punishment:

The teleologic view of punishment is that any punishment should always be accompanied by some beneficial effect. Even though the act of punishing someone may itself be considered “evil”, the overall punishment will be considered “good” if it provides some form of social benefit to the society. For example, if a criminal is punished for a crime and this punishment serves to dissuade potential criminals from committing future crimes, then the overall social value of the punishment is positive.[3]

Punishment from this perspective provides a good model for computer systems as any criminal act will be handled such that punishment is beneficial for the system. A simple example of such a transaction can be visualized through the management of bandwidth. If there is a particular computer who is deemed to be a criminal bandwidth hog, using more resources than it is allowed, the perpetrator’s network connection may be throttled. This punishment would correct the deviant computer while free the resources for other computers in the system to utilize.

Retributive View of Punishment:

Retributive punishment is defined by the belief that punishment itself is either just or intrinsically valuable even if there is social benefit to the punishment. This view is probably best characterised by the phrase “eye for an eye”. Essentially punishment is dispensed because it is necessary to inflict harm on those who do bad things but society may not get any benefit from the punishment.[3] The point of view of retributive punishment is that it is better to punish someone who commits a crime regardless of the severity of the punishment.[2]

It is also important to discuss retributive punishment in comparison to retaliation. Although they both incorporate the concept punishment as a just, and necessary, act, they have very different goals. Retribution focuses on the wrongdoing of the criminal whereas retaliation is based on the right of the victim to seek punishment. Retaliation is based on the concept of deterrence; if you are convicted of a crime then someone will get to exact revenge and thus you will pay a price. Retribution requires a criminal to pay a price for the crime committed and thus he should internalize how the crime has a negative effect on society. [1]

Although the necessity of punishment is certainly true, it is hard to see a situation where truly retributive punishment is beneficial for a computer system. Since computers on a distributed system are sharing finite resources, inflicting punishment on criminal computers, without considering the benefit/harm to the society, may result in further negative effects on the system. If a crime has been committed, the effects of that crime have already had an effect on the computing system. Punishing the perpetrator of the criminal act will not reverse the effect of the act, and it may adversely affect the system.

For example, suppose a computer is caught conducting spam attacks on other computers and the punishment for this act is to remove the computer from the network. It may be the case that the criminal computer had previously provided a very efficient connection to some data set but now there is no way to communicate with this computer. As a result, members of the remaining network must use a less efficient connection to reach the same data, thus the punishment had a negative effect on the system.

Teleologic Retributive View of Punishment:

This third view of punishment combines the concept of the need to punish within the limits of what is considered reasonable punishment for the crime. From this perspective, punishment is necessary and provides a valuable service to society but it is only enforced within acceptable limits.

To illustrate this view of punishment, consider the spam attack example from the retributive section above. If the punishment is reduced from being removed from the network to simply blocking a specific type of communication originating from the criminal computer, then it would be considered a teleological retributive punishment. This new punishment would match the severity of the crime but also still allows the other computers on the network to utilize the efficient network path through the criminal computer.

Structure of Punishment

To maintain a stable and efficient distributed system, punishment requires structure, or more accurately, there needs to be some power imbalance designed within the system such that some computers can hand out punishments upon other, criminal computers. Here we will briefly discuss a few methods which may be used to implement a penal system into a society.

Sovereign Rule:

In the 1600s, Thomas Hobbes wrote a dissertation on how government and society should be structured. Within this work Hobbes discusses how punishment should be handled by a sovereign ruler. In this system, there is a known set of laws which originate from a single entity which exists above the law. This sovereign ruler is the highest authority of the law, but he may assign lesser judges who may carry out punishment in accordance to the laws.[4]

In this system, breaking the law is never excusable as the law is known to all members of the society. The exception to this are any members of society that are without reason, for example “children and madmen”. Punishment is a necessary evil and the sovereign has the right to punish any criminal in order to protect the “commonwealth”. The sovereign can even order other subjects to punish criminals but he may not order a criminal to punish himself as this violates the law of self preservation. To balance the system the sovereign may also reward individuals and thus the balance of punishment and reward are the “nerves and joints which move the limbs of a commonwealth.”[4]

Essentially, sovereign rule is one overall leader of justice who determines what is right and wrong in order to best serve the needs of a system.

Corporal Punishment. Economic Punishment, and Prison:

Human punishment commonly falls into three overlapping categories; corporal punishment, economic punishment, and prison. Corporal punishment involves inflicting pain or possibly disfiguring a criminal in response to the crime committed. The main idea is that the criminal should serve as a demonstration of the terrible things that befall those who break the law. Furthermore, any criminal who is disfigured must live with a visual reminder of the act they committed, thus imparting shame upon the perpetrator and allowing others in society to form a conceptual model of the type of person that individual is.[5]

Economic punishment is forcing a criminal to pay a fine for the act committed. The main idea is to make criminals internalize the social costs of the crime they committed. The penalty fine imposed upon the criminal may not be equal to the social cost of the crime committed but it should cause the criminal the same amount of distress as the crime that was committed. [1] Prison is a modern method of punishment by which criminals are forced to exist under the watch of professionals and it is up to the discretion of the professionals as to when the punishment is complete. For example, it is up to lawyers, judges, psychologists and prison guards determine when a criminal’s prison sentence has ended.[5]

These three methods are not mutually exclusive as commonly criminals may be asked to pay a penalty fine as well as serve a prison sentence; however they all serve different purposes. All three punishment types serve as a deterrent to future criminals but each method has a different active agent; corporal punishment uses shame, economic punishment uses monetary handicapping, and prison focuses on reducing personal freedoms. These three concepts may be very useful to a distributed computer justice system.

Addition Concepts Related to Justice

Morality:

For a system of justice to be effective, a known moral code must exist within the society. Friedrich Nietzsche provides one interpretation of morality based on social position which is divided into two categories; “master-morality” and “slave-morality”. Master-morality is split based on good vs. bad, for example, good would be things like wealth, strength, health, and power, while bad is associated terms like poor, weak, sick, and pathetic. Slave-morality, on the other hand, is based on good vs. evil, for example, good would be terms like charity, piety, restraint, meekness, submission, while evil terms are worldly, cruel, selfish, wealthy, and aggressive.[6]

Although some of these terms make no sense in the realm of computers, others certainly could work as a basis for computer morality. For example, if there were measure of strength, health, wealth based on network and data concepts like bandwidth, latency, data integrity, etc. than certain computers could be more "good" than others. Similarly, if computers were acting selfishly, cruelly or aggressively (say DoS or spam attacks) then those computers would be considered as morally "bad". Based on these moral evaluations, computers could have relationships created or destroyed. Moreover, relationship parameters could be given to computers so that if you don't care how something gets accomplished (whether it is morally good or not) then you could tell your computer to allow less moral interactions to occur.

If morality was introduced to a distributed computer system that already has a reliable reputation mechanism, then all computers would be able to know how other computers behave "socially”. This would further allow punishment methods based on shame, to be exacted based on how "bad" a computer’s moral code is. An offending would then have to rebuild a positive moral reputation before it could participate in more trusted social interactions.

Intent:

Mens Rea - state of the mind:

It is said that a crime consists of two elements, the actus reus, and the mens rea. The actus reus defines the action of the crime, and the mens rea defines the mental state. The mental state of a person is highly regarded as being relevant to the punishment of crimes and the Model Penal Code (“MPC”) is used to categorize the mens rea into four levels: purposely, knowingly, recklessly, and negligently. These levels rank from being the highest is acting upon a crime on purpose, and the lowest is being part of a crime negligently. An example of such a case would be when one distinguishes between whether a car hitting someone had been done intentionally, or by accident.[7]

For a computer, there is no such thing as "intent", there is only computation. As such, discovering the intent of a computer is a meaningless task. To handle this issue a system of justice for computers can take one of two approaches; it can attempt to discover the intent of the user of the computer before distribution of punishment, or it can punish all perpetrators of an act with the same severity regardless of intent. If a computer justice system attempts to discover the intent of the human operating a computer then this system must involve a human who can decipher human reason. This would create a justice system that would have a bottleneck at the human investigation point. For our purposes, we propose that all perpetrators of a deviant act are punished with the same severity to prevent the need for human interventions/investigations.

Computer Fraud and Abuse Act:

The digital age has brought on many new kinds of crimes on the Internet in computers. An example of preventative measures for these crimes was created in 1984 by the United States of America’s congress called the Computer Fraud and Abuse Act(“CFAA”). This criminal statute was built under the ideas of mens rea, and the MPC. After being first implemented, there has had to be many changes to it because of unspecific instances of how different crimes were categorized based on the mens rea. The change between “knowingly” and “intentionally” doing an act would change in degrees of punishment as well as accessing a system and damaging a system had to be more specified over time.[8]

One well know case is of Robert Tappan Morris, who was a 1st year graduate student in Cornell University who attempted to demonstrate the inadequacies of current security measures on the computer networks(INTERNET) by releasing a worm virus. The virus had propagated faster than he had intended and attempted to release the instructions on how to kill the worm, but it had been too late, and many computers across the INTERNET had been affected. The government had to try and prove that it was his intent to access unauthorized computers, which he did, and they also tried to prove that it was his intent to damage the machines, but at that point damaging machines had no category in mens rea.[9]

Justice Involving Computers

Applying justice to computers

The first issue arises from the discussion of the mens rea. Some might say that the computer executes commands that is inputted by the user, so this would mean that everything that the computer does, must be on purpose, because it is just following instructions. This may be true except in the following case that an error has occurred in the system, or a bit has gone missing and the address to which sensitive information was to be sent has now been sent to an incorrect address. If this error created many losses to some entity, would the user be blamed for this error or would the computer be blamed for negligently sending to the wrong address? This type of situation seems similar to how humans may be charged for killing someone, the difference between murder and manslaughter is the intent.[12] With the current structure of how computers are set up, it is difficult to map a mens rea scheme to it’s inner workings. (*perhaps if the computer was running some genetic programming to create a program which it deemed good, and then intentionally used it, then that intent to use differs from it continually creating new programs until it decides one is suitable)

Assuming that the state of mind of a computer has been decided, the next thing to consider would be how would one prevent a computer from doing malicious actions. Attempting to follow in the footsteps of the general deterrence theory, we would try to instill some sort of fear of consequences/shame that would come from causing malicious actions. The problem with this approach to justice is difficult because of the idea that computers do not have feelings, and doing any kind of work such as word processing to a denial of service attack would be equal in aspects to what it prefers to do. Deterring possible criminals only works if they are afraid of the consequences and can not accept the ratio of profit over penalty that they will procure from a malicious act. If a punishment for a computer was to execute many functions for a long period of time, the computer itself would not care any less if it were doing those functions or standing by idle. However, for the human that might force it’s computer to do such malicious actions may be deterred from doing so because of the consequences that might follow from the law, or the possible performance drop from his own computer. The penalties set in place are currently only going to affect a human, whether it be sentenced to jail or confiscation of the physical computer, the human aspect of the problem will be removed from the computer element. Had it been that the computer itself was the only one punished for such malicious actions, then it would not prevent further malicious actions from occurring on a computer network by the human user using another computer terminal.[13]

Since a distributed system would want to one day grow to a global scale, the laws and punishments can not be enforced in a legal sense because of jurisdiction issues, therefore, the implementation of a new system must be done so that computers on the system will be deterred from malicious actions. Implementing a morality system that has every node on the system with their personal morality rating will allow for nodes to communicate with other nodes based on how low or high the rating is. From lowering in morality because of malicious actions and raising it by being more helpful to the system will allow for computers to "care" for who they are communicating with, and to also feel shame for when their morality is so low that they can barely communicate with others.(Lowest level might equal to expulsion) Based on this simulated feeling of care and shame, this might allow for a justice system to be implemented onto computers.

Possible Implementation

Designing a complete justice system implementation is far beyond the scope of this project. This does not mean it is not important to describe the important features that a fully functional system would require and outline the potential benefits and shortcomings of the system. In fact, our team could not come up with one complete system that would be feasible; instead we have two potential implementations each with there own positives and negatives. Although the implementations have separate modes of operation, they both take a teleologic-retributive approach to punishment; by this we mean punishment is necessary but we will not inflict punishmnets that will negatively affect the performance or stability of the distributed system.

The remainder of this article will detail the two justice system implementations at a high level and describe how each system could handle three deviant behaviour scenarios; Comment Spam, Denial of Serice Attacks and Phishing.

Local "Justice Web" Implementation

provide details of implementation here.

Overview

The implementation described in the following sections takes into consideration the above discussion involving justice, and applying it to the management of a computer network. The implementation is an incrementally-deployable justice system currently entitled the "Justice Web".

The Justice Web is essentially a computer network that applies a morality rating to those inside the network, as well as connections coming from outside the network. The morality rating is based on previous actions taken by the entity in question, and the rating is decided by the leader(s) of the network. It is up to the individual nodes within the network to enforce restrictions based on the morality rating master list, which is distributed throughout the network.

In order to implement the Justice Web, certain assumptions must be made:

All computers attempting a connection to the network must be able to be authentically identified whether they are inside or outside of the Justice Web.
Probably more.

Morality Rating

In the Justice Web, "Morality Rating" (MR) is assigned by the Judge or Judges of the network. The morality rating determines a user's access to a server, based on that server's rulings. For example, a server within the Justice Web could have a rule that any connections below MR: -100 are not allowed to connect. In addition, users within the Justice Web use their MR to gain access to resources publicly available to those within the network.

Morality ratings are local to each network running the Justice Web, so a computer that has a bad MR in one network might have a good MR in another.

Master List

The "Master List" is a database that stores the morality ratings of a Justice Web, and is maintained by the judges. servers within the network store a "Slave List" which mirrors a subset of the entries in the master list. The subset is dependent on the rules implemented by the server. For instance, if a server decides to deny any connections with an MR: -100, all entries that have MR: -100 would be stored in the slave list.

Judges

A Justice Web needs one or more judges to function. When an offense occurs, a node in the network submits a claim, along with an encrypted evidence log and ID of the offender. If the evidence is sufficient, the judge changes the MR of the offender based on the severity of the offense.

A judge is appointed by default as the node in the network with the highest MR (Usually the owner), though this option is implementation specific. Alternatively, a ring of judges is appointed and given the power to lower the MR of other judges, but the evidence must be validated by a majority of the judges. This would be implemented as an attempt to keep the judges in check.

Rules

Each Justice Web specifies its own set of rules. A rule consists of three parts: the offense, what level of proof is needed, and the severity of the punishment. The creation of these rules is by default left up to the judges. As stated above, the rules are used by the system during the judgement of claims.

Evidence Logs

Evidence logs are stored on each node of the Justice Web, and they keep track of the most recent network activity of that computer. the logs are encrypted to ensure that the computer making the claim did not tamper with the evidence. Upon being retrieved by the Judges, the logs are decrypted and processed. Depending on the network, the evidence made available may or may not be sufficient to justify the claim.

Membership

There are three possible reasons for joining a Justice Web. You could either be:

A server looking for protection from illegitimate traffic and other computers with dispositions of malcontent.
A client looking for protection from phishing attacks and network viruses. (Though this might contradict with the Slave-Master List concept)
A server or client looking for access to distributed resources within the network.

When joining a Justice Web, a computer retains its MR from when it was outside of the network. In other words, you can't reset your score just by joining the network.

Jurisdictions

As a Justice Web grows, it would become infeasible for a ring of judges to handle every single claim. If the network is sufficiently large, it would make sense to implement hierarchy so that the less severe claims are handled by lower rungs. Again, this would be implementation specific, but it would especially make sense in the case of multiple Justice Webs joining or forming some sort of alliance.

Global Implementation

provide details of implementation here.

Use Case Investigation

Case 1: Comment Spam:

The first deviant act we will investigate is comment spam. The type of spam we are focusing on are usually automated scripts which insert the same comment, usually including links to other destination websites, on forums of public sites. Although usually annoying, these comments can direct users to locations where malicious code may be introduced to unsuspecting users who trust the content of the original site where the forum was hosted.

Local Implementation Solution:

provide details

Global Implementation Solution:

provide details

Case 2: Denial of Service:

Denial of service is the act in which a service that is normally available is being accessed by either numerous ip's through ip spoofing or a distributed DoS attack which has multiple valid ip's, which in turn, bog down the system because the service capacity is being constrained. By either maxing out or flooding a service by multiple requests, a system will either be shutdown or very difficult to use, regardless, the outcome is a denial of that service. We have come up with some solutions to how to punish the computers that participate in such an attack, voluntary or not.[14]

Local Implementation Solution:

provide details

Global Implementation Solution:

provide details

Case 3: Phishing:

Phishing provides an interesting challenge for a justice system as the deviant act involves a website that is perfectly legal. A phishing attack occurs when a user visits a website and enters information/clicks on a link expecting the site he/she is visiting to be secure. Unfortunately the site they have been directed to has been designed to imitate the appearance and behaviour of the site the user wishes to visit, but it is not actually the real webpage. As a result, a user may expose private information to parties that are not legally entitled to view the data. Although it is not illegal for the "fake site" to exist, the usage of the site is an act that must be addressed by the justice system.

Local Implementation Solution:

provide details

Global Implementation Solution:

provide details

Resources

[1] Posner, Richard A., Retirbution and Related Concepts of Punishment, The Journal of Legal Studies Vol. 9 No. 1, University of Chicago Press, 1980. PDF

[2] Rawls, John, A Theory of Justice: Revised Edition, Harvard University Press, 2003. PDF (preview copy)

[3] Ezorsky, Gertrude, Philosophical Perspectives on Punishment, State University of New York Press, 1972. HTML (preview copy)

[4] Hobbes, Thomas, The Leviathon, first published 1651, republished by Forgotten Books, 2008. HTML

[5] Foucault, Michel, Discipline & Punish: The Birth of the Prison, Random House, New York, 1995. PDF (preview copy)

[6] Nietzsche, Friedrich, Ecce Homo & The Anarchist translated by Thomas Wayne, New York, 2004. PDF (preview copy)

[7] Haeji Hong, Hacking Through the Computer Fraud and Abuse Act, originally published in 24 U.C. DAVIS L. REV. 283 (1998), HTML (part A)

[8] Haeji Hong, Hacking Through the Computer Fraud and Abuse Act, originally published in 24 U.C. DAVIS L. REV. 283 (1998), HTML (part B)

[9] 928 F. 2d 504 - Court of Appeals, US v. Morris, 2nd Circuit 1991, HTML (case file)

[10] Charles F. Horne(1915),Claude Hermann Walter Johns, The Encyclopaedia Britannica, 11th ed 1910-, Ancient History Sourcebook:Code of Hammurabi, c. 1780 BCE, Translated by L. W. King, Paul Halsall March 1998, HTML (Internet History Sourcebook)

[11] Scott D. Sagan, Review: History, Analogy, and Deterrence Theory, The MIT Press, 1991, HTML (book link)

[12]Rollin M. Perkins, A Rationale of Mens Rea, Harvard Law Review, 1939, HTML (book link)

[13] Marquis Beccaria, Of Crimes and Punishments, Translated by: Edward D. Ingraham, Philip H. Nicklin: A. Walker, 1819, HTML (essay translation)

[14] Roger.M.Needham, Denial of Service,ACM, New York, USA ,1993,PDF