The identification of vulnerabilities in protein networks is a promising approach to predicting potential therapeutic targets. Different methods have been applied to domain-specific applications, with an emphasis on single-node deletions. There is a need to further assess significant associations between vulnerability, functional essentiality and topological features across species, processes and diseases. This requires the development of open, user-friendly systems to generate and test existing hypotheses about the vulnerability of networks in the face of dysfunctional components. We implemented methodologies to estimate the vulnerability of different networks to the dysfunction of different combinations of components, under random and directed attack scenarios. To demonstrate the relevance of our approaches and software, published protein-protein interaction (PPI) networks from S. cerevisiae, E. coli and H. sapiens were analyzed. A PPI network implicated in the development of human heart failure, and signaling networks relevant to Caspase3 and P53 regulation were also investigated. Known essential proteins (individually or in groups) have no detectable effects on network stability. Some of the most vulnerable proteins are neither essential nor hubs. Known diagnostic biomarkers have little effect on the communication efficiency of the disease network. Predictions made on the signaling networks are consistent with recent experimental evidence. Our system, which integrates other quantitative measures, can assist in the identification of potential drug targets and systems-level properties. The system for large-scale analysis of random and directed attacks is freely available, as a Cytoscape plugin, on request from the authors.