Alan McCabe

Neural Network-based Handwritten Signature Verification

Handwritten signatures are considered as the most natural method of authenticating a person’s identity (compared to other biometric and cryptographic forms of authentication). The learning process inherent in Neural Networks (NN) can be applied to the process of verifying handwritten signatures that are electronically captured via a stylus. This paper presents a method for verifying handwritten signatures by using a NN architecture. Various static (e.g., height, slant, etc.) and dynamic (e.g., velocity, pen tip pressure, etc.) signature features are extracted and used to train the NN. Several Network topologies are tested and their accuracy is compared. The resulting system performs reasonably well with an overall error rate of 3:3% being reported for the best case.

Handwritten Signature Verification Using Complementary Statistical Models

This paper describes a system for performing handwritten signature verification using complementary statistical models. The system analyses both the static features of a signature (e.g., shape, slant, size), and its dynamic features (e.g., velocity, pen-tip pressure, timing) to form a judgment about the signer’s identity. This approach’s novelty lies in combining output from existing Neural Network and Hidden Markov Model based signature verification systems to improve the robustness of any specific approach used alone. The system performs reasonably well and achieves an overall error rate of 2:1% in the best case. The results of several other experiments are also presented including using less reference signatures, allowing multiple signing attempts, zero- effort forgery attempts, providing visual feedback, and signing a password rather than a signature.

Markov Model-Based Handwritten Signature Verification

Biometric security devices are now permeating all facets of modern society. All manner of items including passports, drivers licences and laptops now incorporate some form of biometric data and/or authentication device. As handwritten signatures have long been considered the most natural method of verifying ones identity, it makes sense that pervasive computing environments try to capitalise on the use of automated Handwritten Signature Verification systems (HSV). This paper presents a HSV system that is based on a Hidden Markov Model (HMM) approach to representing and verifying the hand signature data. HMMs are naturally suited to modelling flowing entities such as signatures and speech. The resulting HSV system performs reasonably well with an overall error rate of 3.5% being reported in the best case experimental analysis.

Online Payments Using Handwritten Signature Verification

Making payments online is inherently insecure, especially those involving credit cards where a handwritten signature is normally required to be authenticated. This paper describes a system for enhancing the security of online payments using automated handwritten signature verification. Our system combines complementary statistical models to analyse both the static features of a signature (e.g., shape, slant, size), and its dynamic features (e.g., velocity, pen-tip pressure, timing) to form a judgment about the signers identity. This approachs novelty lies in combining output from existing Neural Network and Hidden Markov Model based signature verification systems to improve the robustness of any specific approach used alone. The system can be used to authenticate signatures for online credit card payments using an existing model for remote authentication. The system performs reasonably well and achieves an overall error rate of 2.1% in the best case.

An Artificially Intelligent Sports Tipper

This paper presents a description of an artificially intelligent model for predicting the outcome of particular sporting contests. Sports prediction (or “tipping”) contests are a common pastime among sports fans worldwide. Many participants in these contests have developed their own systems (computerised or otherwise) with which they select winners. The major problems with these systems however is that the user is often swayed by emotion, misguided intuition, or they do not capture enough meaningful information about the competing teams. The work described here is an attempt to extract insightful information from sporting contests in an effort to make objective predictions about likely winners. It is not meant as an aid to gambling, but rather an interesting case study of using neural networks for predicting probabilistic events in a sporting scenario.

Bidding Agents That Perpetrate Auction Fraud

This paper presents a software bidding agent that inserts fake bids on the seller’s behalf to inflate an auction’s price. This behaviour is referred to as shill bidding. Shill bidding is strictly prohibited by online auctioneers, as it defrauds unsuspecting buyers by forcing them to pay more for the item. The malicious bidding agent was constructed to aid in developing shill detection techniques. We have previously documented a simple shill bidding agent that incrementally increases the auction price until it reaches the desired profit target, or it becomes too risky to continue bidding. This paper presents an adaptive shill bidding agent which when used over a series of auctions with substitutable items, can revise its strategy based on bidding behaviour in past auctions. The adaptive agent applies a novel prediction technique referred to as the Extremum Consistency (EC) algorithm, to determine the optimal price to aspire for. The EC algorithm has successfully been used in handwritten signature verification for determining the maximum and minimum values in an input stream. The agent’s ability to inflate the price has been tested in a simulated marketplace and experimental results are presented.

Pre-processing of On-line Signals in Noisy Environments

This paper introduces the extremum consistency (EC) algorithm for avoiding local maxima and minima in a specialised domain. The most notable difference between this approach and others in the literature is that it places a greater importance on the width or consistency of an extremum than on its height or depth (amplitude). Short-term, high amplitude extrema can be encountered in many typical situations (such as noisy environments or due to hardware imprecision) and can cause problems with system accuracy. The EC algorithm is far less susceptible to these situations than hill climbing, convolution, thresholding etc., and tends to produce higher quality results. This paper describes the algorithm and presents results from practical experimentation, which illustrates its superiority over other forms of local extrema avoidance in three real-world applications