abstract: E-Mail header injection vulnerability is a class of vulnerability that can occur in web applications that use user input to construct e-mail messages. E-Mail injection is possible when the mailing script fails to check for the presence of e-mail headers in user input (either form fields or URL parameters). The vulnerability exists in the reference implementation of the built-in “mail” functionality in popular languages like PHP, Java, Python, and Ruby. With the proper injection string, this vulnerability can be exploited to inject additional headers and/or modify existing headers in an e-mail message, allowing an attacker to completely alter the content of the e-mail. This thesis develops a scalable mechanism to automatically detect E-Mail Header Injection vulnerability and uses this mechanism to quantify the prevalence of E- Mail Header Injection vulnerabilities on the Internet. Using a black-box testing approach, the system crawled 21,675,680 URLs to find URLs which contained form fields. 6,794,917 such forms were found by the system, of which 1,132,157 forms contained e-mail fields. The system used this data feed to discern the forms that could be fuzzed with malicious payloads. Amongst the 934,016 forms tested, 52,724 forms were found to be injectable with more malicious payloads. The system tested 46,156 of these and was able to find 496 vulnerable URLs across 222 domains, which proves that the threat is widespread and deserves future research attention. / Dissertation/Thesis / Masters Thesis Computer Science 2016

Today, new web applications are made every single day with increasingly more sensitive data to manage. To ensure that no security vulnerabilities such as data leakage in web applications exist, developers are using tools such as a web vulnerability scanner. This type of tool can detect vulnerabilities by automatically finding input fields where data can be injected and performing different attacks on these fields. One of the most common web vulnerability scanners is OWASP ZAP. Web vulnerability scanners were first developed during a time when traditional multi-page applications were prominent. Nowadays, when modern single-page applications have become the de facto standard, new challenges for web vulnerability scanners have arisen. These problems include identifying dynamically updated web pages. This thesis aims to evaluate the techniques used by OWASP ZAP and several other web vulnerability scanners for identifying two of the most common vulnerabilities, SQL injections and cross-site scripting. This issue is approached by testing the selected web vulnerability scanners on deliberately vulnerable web applications, to assess the performance and techniques used, and to determine if the performance of OWASP ZAP could be improved. If an identified technique in another web vulnerability scanner performed better than the counterpart in OWASP ZAP, it will be implemented in OWASP ZAP and evaluated. From the tests performed, it could be concluded that the performance of OWASP ZAP was lacking in the search for input fields, where a depth-first search algorithm was used. The breadth-first search algorithm used by other scanners was shown to be more effective in specific cases and was therefore implemented in OWASP ZAP. The result shows that the use case for the two algorithms differs between web applications and by using both of the algorithms to find vulnerabilities, better performance is achieved.

Avec le développement croissant d’Internet, les applications Web sont devenues de plus en plus vulnérables et exposées à des attaques malveillantes pouvant porter atteinte à des propriétés essentielles telles que la confidentialité, l’intégrité ou la disponibilité des systèmes d’information. Pour faire face à ces malveillances, il est nécessaire de développer des mécanismes de protection et de test (pare-feu, système de détection d’intrusion, scanner Web, etc.) qui soient efficaces. La question qui se pose est comment évaluer l’efficacité de tels mécanismes et quels moyens peut-on mettre en oeuvre pour analyser leur capacité à détecter correctement des attaques contre les applications web.Dans cette thèse nous proposons une nouvelle méthode, basée sur des techniques de clustering de pages Web, qui permet d’identifier les vulnérabilités à partir de l’analyse selon une approche boîte noire de l’application cible. Chaque vulnérabilité identifiée est réellement exploitée ce qui permet de s’assurer que la vulnérabilité identifiée ne correspond pas à un faux positif. L’approche proposée permet également de mettre en évidence différents scénarios d’attaque potentiels incluant l’exploitation de plusieurs vulnérabilités successives en tenant compte explicitement des dépendances entre les vulnérabilités.Nous nous sommes intéressés plus particulièrement aux vulnérabilités de type injection de code, par exemple les injections SQL. Cette méthode s’est concrétisée par la mise en oeuvre d’un nouveau scanner de vulnérabilités et a été validée expérimentalement sur plusieurs exemples d’applications vulnérables. Nous avons aussi développé une plateforme expérimentale intégrant le nouveau scanner de vulnérabilités, qui est destinée à évaluer l’efficacité de systèmes de détection d’intrusions pour des applications Web dans un contexte qui soit représentatif des menaces auxquelles ces applications seront confrontées en opération. Cette plateforme intègre plusieurs outils qui ont été conçus pour automatiser le plus possible les campagnes d’évaluation. Cette plateforme a été utilisée en particulier pour évaluer deux techniques de détection d’intrusions développées par nos partenaires dans le cadre d’un projet de coopération financé par l’ANR, le projet DALI. / With the increasing development of Internet, Web applications have become increasingly vulnerable and exposed to malicious attacks that could affect essential properties such as confidentiality, integrity or availability of information systems. To cope with these threats, it is necessary to develop efficient security protection mechanisms and testing techniques (firewall, intrusion detection system,Web scanner, etc..). The question that arises is how to evaluate the effectiveness of such mechanisms and what means can be implemented to analyze their ability to correctly detect attacks against Webapplications.This thesis presents a new methodology, based on web pages clustering, that is aimed at identifying the vulnerabilities of a Web application following a black box analysis of the target application. Each identified vulnerability is actually exploited to ensure that the identified vulnerability does not correspond to a false positive. The proposed approach can also highlight different potential attack scenarios including the exploitation of several successive vulnerabilities, taking into account explicitly the dependencies between these vulnerabilities. We have focused in particular on code injection vulnerabilities, such asSQL injections. The proposed method led to the development of a new Web vulnerability scanner and has been validated experimentally based on various vulnerable applications.We have also developed an experimental platform integrating the new web vulnerability scanner, that is aimed at assessing the effectiveness of Web applications intrusion detection systems, in a context that is representative of the threats that such applications face in operation. This platform integrates several tools that are designed to automate as much as possible the evaluation campaigns. It has been used in particular to evaluate the effectiveness of two intrusion detection techniques that have been developed by our partners of the collaborative project DALI, funded by the ANR, the French National Research Agency

Both for its simplicity and efficiency to search for the most critical security vulnerabilities that could exist within a web application, a web vulnerability scanner is a popular tool among any company that develops a web application. With the existence of many different scanners that are available to use, one is unlikely the same as the other and the results attained when evaluating these scanners in relation to each other are often not the same. In this thesis, three different open source web vulnerability scanners are evaluated and analysed based on their ability to find SQL injection and cross-site scripting vulnerabilities. The scanners were used on several open source deliberately broken web applications that acted as benchmarks. The benchmarks that caused much diversity in the results from the scanners were further investigated. When analysing the scanners based on the results, both the actual results were analysed on what caused the diversity but most of all the source code of the scanners were explored and investigated. It could be found that the techniques used by the scanners were essentially similar but contained several minor differences that caused the diversity in the results. Most differences were dependant on the variation of the predefined payloads injected by the scanners, but it could also be found that the approaches used to determine if a vulnerability was detected or not could vary as well. The finalised result concluded in a report that reveals and demonstrates the different approaches that any web vulnerability scanner could use and the limitations of them.

The popularity of the internet has made the use of web applications ubiquitous and essential to the daily lives of people, businesses and governments. Web servers and web applications are commonly used to handle tasks and data that can be critical and highly valuable, making them a very attractive target for attackers and a vector for successful attacks that are aimed at the application layer. Existing misuse and anomaly-based detection and prevention techniques fail to cope with the volume and sophistication of new attacks that are continuously appearing, which suggests that there is a need to provide new additional layers of protection. This work aims to design a new layer of defense based on deception that is employed in the context of web application-layer traffic with the purpose of detecting and preventing attacks. The proposed design is composed of five deception strategies: Deceptive Comments, Deceptive Request Parameters, Deceptive Session Cookies, Deceptive Status Codes and Deceptive JavaScript. The strategies were implemented as a software artifact and their performance evaluated in a testing environment using a custom test script, the OWASP ZAP penetration testing tool and two vulnerable web applications. Deceptive Parameter strategy obtained the best security performance results, followed by Deceptive Comments and Deceptive Status Codes. Deceptive Cookies and Deceptive JavaScript got the poorest security performance results since OWASP ZAP was unable to detect and use deceptive elements generated by these strategies. Operational performance results showed that the deception artifact could successfully be implemented and integrated with existing web applications without changing their source code and adding a low operational overhead.