Umphress and Williams have shown that individual differences in digraph latency may provide a means of accurately verifying the identity of computer users. The present research refined this technique by exploring inter and intra subject differences in digraph latency distributions. Experiment 1 showed that there is marked heterogeneity in the latency with which individual subjects type different digraphs. Consequently, it was found that typist verification accuracy improved when a digraph-specific index of the distance between test and reference digraph latencies was employed. Experiment 1 also showed the utility of nonlinear modelling as a tool to establish optimum verification parameter settings. Experiment 2 showed that the use of a common low-pass temporal filter cutoff setting for all typists when screening digraphs is unwise. It was found that there is a significant interaction between subjects and filter settings such that verification accuracy may improve if subject-specific filter settings are used.

Optimizing digraph-latency based biometric typist verification systems

Structurally object-oriented database systems are a new class of dedicated data storage systems which are intended to form the basis of CAD, CASE, and other design environments which are to support large, distributed development teams. Several concepts of discretionary access controls (DAC) for such systems have been proposed; these concepts support nested complex objects and nested working groups. However, they do not support roles in development teams such as designers, reviewers, managers etc., whose access rights are typically given in terms of object types rather than only in terms of objects. This paper presents a concept of type-level DAC which is intended to complement instance-level DAC and to support roles in development projects. The concept consists of a formal model of the state of the object base with regard to access controls and a formula which derives from this state and the security context of a process the type-rights of this process. Our model has virtually no built-in, enforced policies; it allows users to realize a wide range of application-specific security policies. It supports multiple inheritance; in order to prevent inconsistent rights on types with common subtypes, certain consistency constraints are introduced. The model is group-oriented in that it supports nested working groups and inheritance of rights along group hierarchies. Access to individual types can be explicitly denied. It is implementable in a distributed system; the administration of rights can be fully decentralized.

Computer Security — ESORICS 92

Subliminal channels axe closely related to covert channels and are used to hide secret information. They abuse the communications resource. Subliminal channels can be introduced in many cryptographic systems, and exploit the inherent randomness of the systems. For example, secret information can be hidden in the randomness of the authenticators of an authentication system. Similarly secret information can be hidden in the randomness (of the prover or verifier) of both zero-knowledge proof systems and signature systems.

A Progress Report on Subliminal-Free Channels

[Sim84] introduced the concept of subliminal channel in the context of signature systems. [Des88b] presented a solution against subliminal channels and extended in [Des88a] the solution to abuse-free coin-flipping, abuse-free generation of public keys, and abuse-free zero-knowledge. In this paper we demonstrate that a whole family of systems (generalized Arthur-Merlin games) can be made abuse-free, avoiding the exhaustive approach of [Des88a]. We will hereto formalize the concept of abuse.

/pdf/making-conditionally-secure-cryptosystems-unconditionally-46mwt3efpf.pdf

Making conditionally secure cryptosystems unconditionally abuse-free in a general context

Cryptography and Security in Computing

Recently, Seberry proposed a method for subliminal message transmission over an insecure channel in the case where authentication but not secrecy is required. Here we examine her ideas in some detail, and propose certain changes to the method that would be necessary for implementation. Disciplines Physical Sciences and Mathematics Publication Details Jones, TC and Seberry, J, Authentication without secrecy, Ars Combinatoria, 21A, 1986, 115-121. This journal article is available at Research Online: http://ro.uow.edu.au/infopapers/1021 AUTHENTICATION WITHOUT SECRECY TERRY C. JONES AND JENNIFER SEBERRY ABSTRACT. Re~eDtly, ~berry proposed a method for Bubliminal meB~ tran~missioll over an insecure channel in the ease where a.uthentication but not secrecy is required. Here we examine her ide8.11 in some detail, and propose certain changes to the method that would be necessary for implementation. Re~eDtly, ~berry proposed a method for Bubliminal meB~ tran~missioll over an insecure channel in the ease where a.uthentication but not secrecy is required. Here we examine her ide8.11 in some detail, and propose certain changes to the method that would be necessary for implementation.

/pdf/authentication-without-secrecy-45he9nlm2j.pdf

Authentication without secrecy

Traditionally, users have been authenticated by asking them to provide some form of password. This password has been stored securely in the computer and used to check the identity of the user at various times, such as when they first log on. However such authentication only proves that the challenged user knows the password it doesn't identify the user. This has often been a security problem in time-shared computer installations, when unauthorised users (the proverbial "hackers") have obtained the passwords of valid users and used these to penetrate the site's security. This has led to much work to identify users uniquely by more secure means, such as fingerprints. Such measures all try to identify a user by checking some attribute of the person. In what follows we will discuss various identification schemes based upon work done at the University of Sydney. Disciplines Physical Sciences and Mathematics Publication Details Jones, T, Newberry, M and Seberry, J, User unique identification, in Proceedings of ACSC-12, University of Wollongong, 1989, 163-172. This conference paper is available at Research Online: http://ro.uow.edu.au/infopapers/1039 User Unique Identifcation Terry Jones Mike Newberry Jennifer Seberry Basser Department of Computer Science University Of Sydney Traditionally, users have been authenticated by asking them to provide some form of password. This password has been stored securely in the computer and used to check the identity of the user at various times, such as when they first log on. However such authentication only proves that the challenged user knows the password it doesn't identify the user. This has often been a security problem in time-shared computer installations, .when unauthorised users (the proverbial "hackers") have obtained the passwords of valid users and used these to penetrate the site's security. This has led to much work to identify users uniquely by more secure means, such as fingerprints. Such measures all try to identify a user by checking some attribute of the person. In what follows we will discuss various identification schemes based upon work done at the University of Sydney. 1. Type-Signatures The idea of a type-signature is one such approach, first proposed by Terry Jones [Jones1985]. It incorporates a statistical measure of the typing style of the user. 1.1 Type-signature Password Systems In December 1986 we commenced testing the concept of the type-signature l . Over the next two months several mini password systems, based around this concept. were developed and installed by Mike Newberry on various machines to test them under different conditions (all of these machines were DEC VAX 11/780's running various forms of UNIX). These systems checked both the type-signature and the password when a user attempted to log-on. The type-signatures were allowed to change over time as users became more proficient with their passwords. Facilities were also established for accounts to be owned by more than one person. The results from these test systems were that valid users were able to log-on 89% of the time, while intruders, once they were told an account's password, could log-on only 43% of the time. It should be noted that in a real system the second figure would be smaller as intruders would have no way of knowing whether they they possessed the correct password, until they guessed the correct typing style. These results are further discussed in [Newberry and Seberry 1987). A type-signature system would be proof against the common password attacks. For example, attempts to exhaustively test every possible password~ either manually or by machine, would be unlikely to succeed, as intruders would be unable to determine whether or not they had chosen the correct password, and had used the wrong typesignature, or had simply entered the wrong password. Certainly attempts to exhaustively 1. This was funded by a grant from A TERB.

/pdf/user-unique-identification-5coye4da0n.pdf

Terry C Jones

Papers

Authentication without secrecy

User unique identification