Section: Scientific Foundations
Confidentiality, Integrity and Availability are the three fundamental properties ruling the security of any information system. Data confidentiality has recently become a major concern for individuals as well as for companies and governments. Several kinds of data are threatened: personal data gathered by visited Web sites or by smart objects used in our daily life, corporate or administrative data stored in piracy-prone servers or hosted by untrusted Database Service Providers. The CSI/FBI reports that database attacks constitute the first source of cyber-criminology and that more than fifty percents of the attacks are conducted by insiders  . In this context, governments are setting up more constraining legislations. The problem is then to translate law statements into technological means: authentication mechanisms, data and communication encryption protocols, access control models, intrusion detection systems, data and operation anonymization principles, privacy preserving data mining algorithms, etc. The area of investigation is extremely large. Our own research program focuses on data access, usage and retention control and on the way this control can be made secure (i.e., tamper-resistant).
Access control management has been deeply studied for decades. Different models have been proposed to declare and administer access control policies, like DAC, MAC, RBAC, TMAC, OrBAC  . While access control management in relational databases is now well established and normalized, new access control models have to be defined to cope with more complex data (e.g., hierarchical and semi-structured data like XML) and new forms of data distribution (e.g., selective data dissemination). Privacy models are also emerging today  . Privacy distinguishes from confidentiality is the sense that the data to be protected is personal. Hence, the user's consent must be reflected in the access control policies and not only the access but also the usage of the data as well as its retention period are safeguarded by law and must be controlled carefully.
Securing the access control against different forms of tampering is a very important issue. Server-enforced access control is widely accepted  but remains inoperative against insider attacks. Several attempts have been made to strengthen server-based security with database encryption   . However, the Database Administrator (or an intruder usurping her identity) has enough privilege to tamper the encryption mechanism and get the clear-text data. Client-based security approaches have been recently investigated. Encryption and decryption occur at the client side to prevent any disclosure of clear-text data at the server. Storage Service Providers proposing encrypted backups for personal data are crude representative of this approach. The management of SQL queries over encrypted data complements well this approach  . Client-based decryption is also used in the field of selective data dissemination (e.g., Digital Right Management). However, the sharing scenarios among users are generally coarse grain and static (i.e., pre-compiled at encryption time). Tamper-resistant hardware can help devising secured database architectures alleviating this problem. Finally, securing the usage of authorized data is becoming as important as securing the access control as far as privacy preservation is concerned. Thus, database encryption, tamper-resistant hardware and their relationships with access control and usage control constitute a tremendous field of investigation.