Today’s software systems tend to be long-living and often process security-critical data, so keeping up with ever-changing security measures, attacks, and mitigations is critical to maintaining their security. While it has become common practice to consider security aspects during the design of a system, OWASP still identifies insecure design as one of the top 10 threats to security. Furthermore, even if the planned design is secure, verifying that the planned security assumptions hold at run-time and investigating any violations that may have occurred is cumbersome. In particular, the configuration of run-time monitors such as the Java Security Manager, which could enforce design-time security assumptions, is non-trivial and therefore used in practice rarely. To address these challenges, we present UMLsecRT for automatically supporting model-based security engineering with run-time monitoring of design-time security specifications and round-trip engineering for propagating run-time observations to the design level. Following the established security-by-design approach UMLsec, security experts annotate system models with security properties that UMLsecRT automatically synchronizes with corresponding source code annotations for the automatic configuration of UMLsecRT’s run-time monitor. To this end, UMLecRT monitors these security properties at run-time without additional effort to specify monitoring policies. Developers can define mitigations for attacks detected at run-time in advance by adjusting the automatically synchronized annotations at implementation time. Triggered by a security violation, UMLsecRT can adapt the design-time models based on run-time findings to facilitate the investigation of security violations. We evaluated UMLsecRT concerning its effectiveness and applicability to security violations extracted from real-world attacks and the DaCapo benchmark, conducted user studies on the usability of the adapted models and the feasibility of UMLsecRT in practice, especially concerning countermeasures, and investigated the scalability of UMLsecRT. To study the applicability of the whole development process, we applied UMLsecRT in two case studies to the Eclipse Secure Storage and the electronic health record system iTrust.

Security is an important quality aspect for modern information systems. Security properties may however be violated if the information system operates in an evolving environment. Environmental changes then trigger reactions which lead to co-evolution of the security design and the corresponding system model. However, updating the security design manually is time-consuming and error-prone. We present an approach to support semi-automatic system co-evolution which responds to environmental knowledge evolution, using the UML security extension UMLsec and graph transformation. The aim is to enable software engineers to react more reliably and effectively to environmental changes and to ensure lifelong compliance of information systems. To evaluate our approach, we conducted a case study on the open-source project iTrust.

Security is an increasingly important quality facet in modern information systems and needs to be retained. Due to a constantly changing environment, long-living software systems “age” not by wearing out, but by failing to keep up-to-date with their environment. The problem is that requirements engineers usually do not have a complete overview of the security-related knowledge necessary to retain security of long-living software systems. This includes security standards, principles and guidelines as well as reported security incidents. In this paper, we focus on the identification of known vulnerabilities (and their variations) in natural-language requirements by leveraging security knowledge. For this purpose, we present an integrative security knowledge model and a heuristic method to detect vulnerabilities in requirements based on reported security incidents. To support knowledge evolution, we further propose a method based on natural language analysis to refine and to adapt security knowledge. Our evaluation indicates that the proposed assessment approach detects vulnerable requirements more reliable than other methods (Bayes, SVM, k-NN). Thus, requirements engineers can react faster and more effectively to a changing environment that has an impact on the desired security level of the information system.