Some ideas for a Master thesis
Interested candidates should contact Nicola Zannone (email: n.zannone * tue.nl (replace * with @) )
Privacy is becoming a urgent issue in information systems nowadays because of the stringent requirements imposed by data protection regulations.
Traditional security approaches based on access control and authorization are not adequate to address these requirements.
The underlying fundamental problem is that those approaches are preventive and thus they are not able to deal with exceptions.
The goal of the project is to develop a practical privacy framework that shifts the problem of preventing infringements into a problem
of detecting infringements.
The project requires investigation on systematic log auditing techniques, use of patterns and privacy metrics to detect and quantify infringements.
Policy compliance over incomplete logs:
Organizations that collect, use and share personal data have to ensure that data are processed according to security and privacy policies.
Compliance is usually assessed by analyzing audit logs, which record user behavior, against policy specifications.
However, the logs maintained by an organization may be incomplete, that is, they may not contain all information needed to determine whether policies have been violated or not.
The goal of this project is to investigate methods and techniques aims to assist auditors in assessing policy compliance over incomplete logs.
Risk/benefit- based infringement management:
In the last years, businesses have become more dynamic and, consequently, organizations have to adapt quickly to changes in structure, businesses, and environmental conditions.
Organizations thus have to deal increasingly often with unexpected situations and exceptions. For instance, a physician can take actions that diverge from the procedures adopted by the hospital to face emergency situations. Preventing such actions may be critical for the life of the patient.
Traditional access control systems are too inflexible to be used in dynamic environments. Their inflexibility often forces users to bypass them and, in extreme cases, to switch off security measures.
The objective of this project is to design and develop an automated infringement management mechanism that supports the system in making informed decision on data access by performing real-time risk/benefit analysis.
Collaborative Access Control:
In recent years, collaboration has become an important aspect of information systems.
With more and more users being connected through the Internet, the need of online collaborative environments has arisen.
This has spurred the development of several collaborative systems like web application platforms (e.g., Microsoft Sharepoint) and social networks (e.g., Facebook and Google+).
Collaborative systems provide users an environment in which they can work together to achieve a common goal.
A main feature of collaborative systems is the possibility to share (possibly sensitive) information.
Indeed, sharing information in a collaboration might be required to achieve a common goal.
This information, however, might be sensitive and should be accessed exclusively by authorized users.
Information is usually protected by means of access control.
However, access control mechanisms usually assume that information are "owned" by single users.
This may not be the case in collaborative environments.
Here, more that one user may retain some rights on the information; moreover, each user may regulate access to information in different capacity.
The goal of this project is to investigate novel access control model for collaborative systems.
Policy administration in TBA:
Tag-Based Authorization (TBA) is a hybrid access control model that combines the ease of use of extensional access control models with the expressivity of logic-based formalisms.
In particular, it allows the users of a system to assign descriptive tags to the system's resources (i.e., subjects and objects); security experts then
write logical authorization policies in terms of those tags. The main limitation of the TBA model is that it lacks support for policy administration.
More precisely, it does not allow policy-writers to specify administrative policies that constrain the tags that users can assign, and to verify the compliance of assigned tags with these policies.
The goal of this project is to enable policy administration in TBA.
Classification of attacks on DBMS:
Anomaly Detection Systems (ADS) can be broadly classified as rule-based or behavior-based. Rule-based approaches detect attacks that match specific malicious patterns, thus detect known attacks only. The behavior-based approaches are based on two phases: a training phase where profiles of normal behavior are created, and a detection phase where new queries are evaluated against the profile. Alerts are raised if the queries do not match the behavior pattern. The advantage of behavior-based ADS is that they detect unknown instances of attacks. The disadvantage is that they don't offer much information about the root cause of the attack. The goals of the project are:
formalize the patterns of SQL attacks
- find relevant features that characterize types (classes) of attacks
- find correlation between features that characterize types (classes) of attacks
- create a SQL attack classifier
- Elisa Bertino and Gabriel Ghinita. 2011. Towards mechanisms for detection and prevention of data exfiltration by insiders: keynote talk paper. In Proceedings of the 6th ACM Symposium on Information, Computer and Communications Security (ASIACCS '11). ACM
- Ashish Kamra, Evimaria Terzi, and Elisa Bertino. 2008. Detecting anomalous access patterns in relational databases. The VLDB Journal 17, 5 (August 2008)
- Elisa Costante, Jerry den Hartog, Milan Petkovic, Sandro Etalle, Mykola Pechenizkiy: Hunting the Unknown - White-Box Database Leakage Detection. DBSec 2014: 243-259