Grid Security

4 Grid Security 4.1 INTRODUCTION In general, IT security is concerned with ensuring that critical information and the associated infrastructures are not compro- mised or put at risk by external agents. Here, the external agent might be anyone that is not authorized to access the aforemen- tioned critical information or infrastructure. The critical infras- tructure we are referring to is that which supports banking and financial institutions, information and communication systems, energy, transportation and other vital human services. The Grid is increasingly being taken up and used by all sectors of busi- ness, industry, academia and the government as the middle- ware infrastructure of choice. This means that Grid security is a vital aspect of its overall architecture if it is to be used for critical infrastructures. A number of observations have been made on critical infras- tructures [1]. It is clear that in today’s world they are highly interdependent, both physically and in their reliance on national information infrastructure. Most critical infrastructures are largely owned by the private sector, where there tends to be a reluctance to invest in long-term and high-risk security-related technologies. Ongoing changes to business patterns are reducing the level of tolerance to errors in these infrastructures. However, there is insuf- ficient awareness of critical infrastructure issues. The growth of The Grid: Core Technologies Maozhen Li and Mark Baker © 2005 John Wiley & Sons, Ltd 124 GRID SECURITY IT and the Internet can therefore have major implications for the economic and the military security of the world. IT infrastructures are changing at a staggering rate. Their scale and complexity are becoming ever greater in scope and func- tional sophistication. Boundaries between computer systems are becoming indistinct; increasingly every device is networked, so the infrastructure is becoming a heterogeneous sea of components with a blurred human/device boundary. There is continuous and incre- mental development and deployment; systems evolve by adding new features and greater functionality at an unremitting pace. These systems are becoming capable of dynamic self-configuration and adaptation, where systems respond to changing circumstances and conditions of their environment [2]. Increasingly there are mul- tiple innovative types of networked architectures and strategies for sharing resources. This obviously leaves gaps for a multiplicity of fault types and openings for malicious faults, as well as attacks from internal and external parties. The actors that may want to compromise critical information or infrastructures are many and varied. They include those that pose national security threats, such as information warriors or agents involved in national intelligence. Alternatively the actors could be terrorists involved in industrial espionage or organized crime and who pose a shared threat to a country. Or the threats could just be local and come from institutional or recreational hackers intent on thrill, challenge or prestige. 4.2 A BRIEF SECURITY PRIMER The goals of security are threefold [2]: first, prevention – prevent attackers from violating security policy; secondly detection – detect attackers’ violation of security policy; finally recovery – stop an attack, assess and repair damage, and continue to function cor- rectly even if the attack succeeds. Obviously prevention is the ideal scenario. In this case there would be no successful attacks. Detection occurs only after someone violates the security policy. It is important when a vio- lation has occurred or is underway that it is reported swiftly. The system must then respond appropriately. Recovery means that the system continues to function correctly, possibly after a period of degraded operation. Such systems are said to be intrusion tolerant. 4.2 A BRIEF SECURITY PRIMER 125 This is very difficult to do correctly. Usually, recovery means that the attack is stopped and the system fixed. This may involve shut- ting down the system for some time, or making it unavailable to all users except those fixing the problem, and then the system resumes correct operation. The three classic security concerns of information security deal principally with data, and are: • Confidentiality: Data is only available to those who are autho- rized; • Integrity: Data is not changed except by controlled processes; • Availability: Data is available when required. Confidentiality is aimed at different issues. The content of a packet during a communication has to be secure to prevent malicious users from stealing the data. In order to prevent unauthorized users retrieving secret information, a common approach is to encrypt data from the sender before sending to the receiver. On the receiving end, the receiver can extract the original information by decrypting the encrypted text. Hence, confidentiality of data transmission is closely related to application of different encryp- tion algorithms. Integrity is the protection of data from modification by unau- thorized users. This is not the same as data confidentiality. Data integrity requires that no unauthorized users can change or modify the data concerned. For example, you want to broadcast a message to the public, which is definitely not confidential to anyone. You have to ensure the data integrity of your message from modifi- cation by unauthorized people. In this instance, you may have to stamp or add your signature to certify the message. The term “availability” addresses the degree to which a system, sub-system or equipment is operable and in a usable state. Additional concerns deal more with people and their actions: • Authentication: Ensuring that users are who they say they are; • Authorization: Making a decision about who may access data or a service; • Assurance: Being confident that the security system functions correctly; 126 GRID SECURITY • Non-repudiation: Ensuring that a user cannot deny an action; • Auditability: Tracking what a user did to data or a service. Authentication implies ensuring the right user executes the granted services or the origin of the data was from the real sender. There are a large number of techniques that may be used to authenticate a user – passwords, biometric techniques, smart cards or certificates. Before starting services between a server and client, there should be a mechanism to identify the privilege of the user. You may consider a user name and password for a logon scheme to be the most common authentication scheme. For example, you have to input your Visa Card password at an ATM terminal or enter a password to gain access to a Web portal. Hence, authenticity is mainly related to the identification of authorized users. Authorization is often a first step towards providing the service of authentication. Authorization enables the decision to allow a particular operation when a request to perform the operation is received. Authorization in existing systems is usually based on information local to the server. This information may be present in Access Control Lists (ACL) associated with files or directo- ries. ACLs are files listing individuals authorized to login to an account (e.g. the UNIX .rhosts file), configuration files nam- ing authorized users of a node and sometimes files read over the network. When applied to distributed systems, authorization mechanisms are required to determine whether a particular task should be run on the current node when requested by a particular principal. Many applications, and in particular applications using distributed systems, can benefit from an authorization mechanism that supports delegation. Delegation is a means by which a user or process authorized to perform an operation can grant the author- ity to perform that operation to another process. Delegation can be used to implement distributed authorization where, for example, a resource manager might allocate a node to a job and delegate authority to use that node to the job’s initiator. Assurance is the counterpart to authorization. Authorization mechanisms allow the provider of a service to decide whether to perform an operation on behalf of the requester of the service. Assurance mechanisms allow the requester of a ser- vice to decide whether a candidate service provider meets the requesters’ requirements for security, trustworthiness, reliability or other characteristics. Assurance mechanisms can be implemented 4.3 CRYPTOGRAPHY 127 through certificates (see Section 4.3.5 for a discussion of certificates) signed by a third party trusted to endorse, license or insure a service provider; certificates are checked as a client selects the providers to contact for particular operations. Non-repudiation is the concept of ensuring that a contract, espe- cially one agreed to via the Internet, cannot later be denied by one of the parties involved. With regard to digital security, non- repudiation means that it can be verified that the sender and the recipient were, in fact, the parties who claimed to send or receive the message, respectively. Auditability is about keeping track of what is happening on a system. The idea is that if there is an intrusion, then the sys- tem operator can find out exactly what has been done and in whose name. Other security concerns relate to: • Trust: People can justifiably rely on computer-based systems to perform critical functions securely, and on systems to process, store and communicate sensitive information securely; • Reliability: The system does what you want, when you want it to; • Privacy: Within certain limits, no one should know who you are or what you do. 4.3 CRYPTOGRAPHY 4.3.1 Introduction Cryptography is the most commonly used means of providing security; it can be used to address four goals: • Message confidentiality: Only an authorized recipient is able to extract the contents of a message from its encrypted form; • Message integrity: The recipient should be able to determine if the message has been altered during transmission; • Sender authentication: The recipient can identify the sender, and verify that the purported sender did send the message; • Sender non-repudiation: The sender cannot deny sending the message. 128 GRID SECURITY Obviously, not all cryptographic systems (or algorithms) realize, nor intend to, achieve all of these goals. 4.3.2 Symmetric cryptosystems Using symmetric (conventional) cryptosystems, data is trans- formed (encrypted) using an encrypted key and scrambled in such a way that it can only be unscrambled (decrypted) by a symmetric transformation using the same encryption key. Besides protecting the confidentiality of data, encryption also protects data integrity. Knowledge of the encryption key is required to produce cipher text that will yield a predictable value when decrypted. Therefore modification of the data by someone who does not know the key can be detected by attaching a checksum before encryption, and verified after decryption. A sketch of symmetric key cryptography is shown in Figure 4.1. 4.3.2.1 Example: Data Encryption Standard (DES) DES consists of two components – an algorithm and a key. The DES algorithm involves a number of iterations of a simple transfor- mation which uses both transposition and substitution techniques applied alternately. DES is a so-called private-key cipher; here data is encrypted and decrypted with the same key. Both sender and receiver must keep the key a secret from others. The DES algo- rithm is publicly known, learning the encryption key would allow an encrypted message to be read by anyone. Encrypt with secret key Decrypt with secret key Plaintext The Internet Cipher text Plaintext Figure 4.1 Symmetric key cryptography 4.3 CRYPTOGRAPHY 129 4.3.3 Asymmetric cryptosystems In asymmetric cryptography, encryption and decryption are per- formed using a pair of keys such that knowledge of one key does not provide knowledge of the other key in the pair. One key, called the public key, is published, and the other key, called the private key, is kept private. The main advantage of asymmetric cryptogra- phy is that secrecy is not needed for the public key. The public key can be published rather like a telephone number. However only someone in possession of the private key can perform decryption. A sketch of asymmetric key cryptography is shown in Figure 4.2. 4.3.3.1 Example: RSA An example of a public-key cryptosystem is RSA, named after its developers, Rivest, Shamir and Adleman (RSA), who invented the algorithm at MIT in 1978. RSA provides authentication, as well as encryption, and uses two keys: a private key and a public key. With RSA, there is no distinction between the function of a user’s public and private keys. A key can be used as either the public or the private key. The keys for the RSA algorithm are generated mathematically, in part, by combining prime numbers. The security of the RSA algorithm, and others similar to it, depends on the use of very large numbers (RSA uses 256 or 512 bit keys). With both symmetric and asymmetric systems there is a need to secure the private key. The private key must be kept private. It should not be sent to others, and it should not be stored on a system where others may be able to find and use it. The stored keys should always be password protected. Another issue, with Encrypt with public key Decrypt with private key Plaintext The Internet Cipher text Plaintext Figure 4.2 Asymmetric key cryptography 130 GRID SECURITY key-based systems, is that the algorithms that are used are public. This means that algorithms could be coded and used to decrypt a message via a brute force method of trying all the possible keys. Such a program would fortunately need a significant amount of computational power to accomplish such a process. With keys of sufficient length, the time to decode a message would be unrea- sonable. Currently, keys are typically 1024–2048 bits in length. However, the availability of cheap computational power, in the form of clusters of PCs, and their doubling in speed every eighteen months, means that the brute force approach may be viable in the future. 4.3.4 Digital signatures Integrity is guaranteed in public-key systems by using digital sig- natures, which are a method of authenticating digital information, in the same manner that an individual would sign a paper docu- ment to authenticate it. A digital signature is itself a sequence of bits conforming to one of a number of standards. Most digital signatures rely on public-key cryptography to work. Consider a scenario, where someone wants to send a message to another and prove its originator, but does not care whether any- body else reads it. In this case, they send an encrypted copy of the message, along with a copy of the message encrypted with their private (not public) key. A recipient can then check whether the message really came from originator by unscrambling the scram- bled message with the sender’s public key and comparing it with the unscrambled version. If they match, the message was really from the originator, because the private key was needed to create the encrypted copy and no one but the originator has it. Often, a cryptographically strong hash function [4] is applied to the mes- sage, and the resulting message digest is encrypted instead of the entire message, which makes the signature significantly shorter than the message and saves considerable time since hashing is generally much faster, byte for byte, than public-key encryption. 4.3.5 Public-key certificate A public-key certificate is a file that contains a public key, together with identity information, such as a person’s name, all of which is 4.3 CRYPTOGRAPHY 131 signed by a Certification Authority (CA). The CA is a guarantor who verifies that the public key belongs to the named entity. Certificates are required for the large-scale use of public-key cryptography, since anybody can create a public–private-key pair. So in principle, if the originator is sending private information encrypted with the recipient’s public key, a malicious user can fool the originator into using their public key, and so get access to the information, since it knows its corresponding private key. But if the originator only trusts public keys that have been signed (“certified”) by an authority, then this type of attack can be pre- vented. In large-scale deployments one user may not be familiar with another’s certification authority (perhaps they each have a different company CA), so a certificate may also include a CA’s public key signed by a higher level CA, which is more widely recognized. This process can lead to a hierarchy of certificates and complex graphs representing trust relations. Public Key Infrastructure (PKI) refers to the software that man- ages certificates in a large-scale setting. In X.509 PKI systems, the hierarchy of certificates is always a top-down tree, with a root certificate at the top, representing a CA that is so well known it does not need to be authenticated. A certificate may be revoked if it is known that the related private key has been exposed. In this circumstance, one needs to look up the Certificate Revoca- tion List, which is often stored remotely, and updated frequently. A certificate typically includes: • The public key being signed; • A name, which can refer to a person, a computer or an organi- zation; • A validity period; • The location (URL) of a revocation list. The most common certificate standard is the ITU-T X.509 [5]. An X.509 certificate is generally a plaintext file that includes informa- tion in a specific syntax: • Subject: This is the name of the user; • Subject’s public key: This includes the key itself, and other infor- mation such as the algorithm used to generate the public key; • Issuer’s subject: CA’s distinguished name (Table 4.1); 132 GRID SECURITY Table 4.1 Distinguished Names (DN) Names in X.509 certificates are not encoded simply as common names, such as “Mark Baker” or “Certificate Authority XYZ” or “System Administrator”. Names are encoded as distinguished names, which are name–value pairs. An example of typical distinguished name attributes is shown below. OU = Portsmouth, L = DSG, CN = Mark Baker A DN can have several different attributes, and the most common are the following: OU: Organizational Unit L: Location CN: Common Name (usually the user’s name). • Digital signature: The certificate includes a digital signature of all the information in the certificate. This digital signature is generated using the CA’s private key. To verify the digital sig- nature, we need the CA’s public key, which is found in the CA’s certificate. 4.3.6 Certification Authority (CA) The CA exists to provide entities with a trustable digital identity which they can use to access resources in a secure way. Its role is to issue (create and sign) certificates, make the valid certificates publicly accessible, revoke certificates when necessary and regu- larly issue revocation lists. The CA must also keep records of all its transactions. A CA can issue personal certificates to users; the purpose of the certificate is to allow users to identify themselves to remote entities. A personal certificate can also be used for digital signatures. A CA can also issue host (server) and service certificates. Each host and service connected to a network must be able to identify itself. Some CAs can issue certificates which validate the identity of subordinate CAs. Some CAs are therefore subordinate CAs and some simply appoint themselves. In either case the CA publishes a document called the Certificate Policy Statement (CPS). This is a document which details the conditions under which it issues [...]... of a firewall can render it valueless as a security tool 4.4 GRID SECURITY 4.4.1 The Grid Security Infrastructure (GSI) Grid security is based on what is known as the Grid Security Infrastructure (GSI) (Figure 4.3), which is now a Global Grid Forum (GGF) standard [6, 7] GSI is a set of tools, libraries, and protocols used in Globus (see Chapter 5), and other grid middleware, to allow users and applications... the gridmap file The gridmap file is used by GSI after authenticating a user A gridmap file at each site specifies grid- id to local-id mapping GSI first checks to see if the user’s grid identity is listed in the gridmap file If the user is not listed, they are denied access to the resource An example of a gridmap file entry is: "/C=UK/O=eScience/OU=Portsmouth/L=DSG/ CN=Mark Baker" mab The Globus command grid_ mapfile_add_user... or declines an incoming security request • None: This is the simplest type of authorization No authorization will be performed • Self: A client will be allowed to use a grid service if the client’s identity is the same as the service’s identity • Gridmap: The gridmap file contains a list of authorized users; this is similar to an ACL Here only the users listed in the service’s gridmap file may invoke... GLOBUS_TCP_PORT_RANGE environment variable to a given port range Grid service Static ports for well-known Grid services These ports are, for example: • 22/TCP (GSI-enabled OpenSSH) • 2119/TCP (Globus Gatekeeper) • 2135/TCP (MDS) • 2811/TCP (GridFTP server) 4.6 POSSIBLE VULNERABILITIES Security mechanisms are not fool proof There always tends to be tension between security and usability So, there will always be vulnerabilities,... e-Science report, A Rough Guide to Grid Security [12], especially note the social engineering techniques that can be used to obtain user credentials 4.6.2 Proxies The proxy mechanism is useful as it provides local authorization without the need to contact the user site for every remote action 150 GRID SECURITY or service access There are some drawbacks that compromise security: • The proxy’s private key... Maheswaran, M., Towards Trust-Aware Resource Management in Grid Computing Systems, Cluster Computing and the Grid 2nd IEEE/ACM International Symposium CCGrid 2002, pp 452–457 Butler, R., Welch, V., Engert, D., Foster, I., Tuecke, S., Volmer, I and Kesselman, C., A National-Scale Authentication Infrastructure, Computer, 33 (12), December 2000, pp 60–66 152 GRID SECURITY Butt, A., Adabala, S., Kapadia, N., Figueiredo,... Control for Securing Shared Resources in Computational Grids, Parallel and Distributed Processing Symposium, Proceedings International, IPDPS 2002, Abstracts and CD-ROM, 2002, pp 206–213 Humphrey, M and Thompson, M., Security Implications of Typical Grid Computing Usage Scenarios, Security Working Group GRIP forum draft, October 2000 Internet Security Glossary, ftp://ftp.isi.edu/in-notes/rfc2828.txt... Group, http://forge.gridforum.org/projects/gsi-wg [7] Foster, I., Kesselman, C., Tsudik, G and Tuecke, S., A Security Architecture for Computational Grids Proceedings of 5th ACM Conference on Computer and Communications Security Conference, 1998, pp 83–92 [8] SSL, http://docs.sun.com/source/816-6156-10/contents.htm [9] OpenSSL, http://www.openssl.org/ [10] Key Conversion, http://www .grid- support.ac.uk/ca/user-documentation/... as will be seen when security is mentioned The chapter then discussed the GSI, and the various authorization modes used in GSI We then put together all the security components previously mentioned and described how a user would get a certificate; manage credentials and transverse a firewall Finally we briefly highlighted possible vulnerabilities that possibly still exist in Grid security 4.8 ACKNOWLEDGEMENTS... delegation (GSI extensions) for secure single sign-on Proxies and delegation PKI for credentials PKI (CAs and certificates) SSL / TLS Figure 4.3 The Grid Security Infrastructure Secure Sockets Layer (SSL) for authentication and message protection 4.4 GRID SECURITY 135 based on a Public Key Infrastructure, with certificate authorities and X.509 certificates GSI provides: • A public-key system; • Mutual . tool. 4.4 GRID SECURITY 4.4.1 The Grid Security Infrastructure (GSI) Grid security is based on what is known as the Grid Security Infrastructure (GSI) (Figure. computer security in general. Errors or small mistakes in the configuration of a firewall can render it valueless as a security tool. 4.4 GRID SECURITY 4.4.1