Teaching computer security is one of the most challenging tasks in computer science, because of the need to successfully integrate abstract concepts and practical applications. Several e-learning systems have been developed to address this issue. However, they usually provide the same material in the same sequence irrespective of the characteristics of the students, such as their knowledge level and learning style. In this paper, an approach to learning style adaptivity is proposed for the teaching of computer security. An e-learning system was developed to provide more personalised and adaptive learning, based on the information perception style of the Felder-Silverman model. This is the dimension of learning style, which has received the least attention in published research. In the approach, a personalised sequence of learning material is generated based on an individual learning style. The approach is evaluated in order to determine its effectiveness in learning provision. An experiment conducted with sixty subjects produced significant results. They indicate that matching computer security learning material, according to the learning style of the students, yields significantly better learning gain and student satisfaction than without matching.