In recent years, several innovative security technologies have been developed. However, many of the novel sensing technologies (e.g. video analytics) do not always feature a high level of reliability. Very often, they need to be precisely tuned to fit specific installations and provide acceptable results. Furthermore, in large installations the number of surveillance operators is low with respect to the number of sensing devices, and operators' tasks include facing critical events, possibly including strategic terrorist attacks. In such human-in-the-loop systems, ergonomics and usability issues need to be carefully addressed to increase system performance in terms of detection probability and low rate of false/nuisance alarms. This paper describes a multi-sensor event correlation approach for augmenting the capabilities of distributed surveillance systems. The aim is to provide advanced early warning, situation awareness and decision support features. The effectiveness of the framework is proved considering threat scenarios of public transportation systems.