Content-based network filtering of encrypted image data

The proliferation of multimedia encryption techniques allows securing various applications including tele-browsing and visio-conferencing. However, these techniques may also constitute useful tools for malicious users to transmit prohibited data without being detected by preventive and reactive security mechanisms. This paper proposes an application-layer filtering strategy for progressively enciphered JPEG2000 image data. It first consists in analyzing the JPEG2000 syntax and second employs progressive decryption and correlation to look for a match between the transmitted flow and a set of pre-defined prohibited patterns. Simulations show that these strategies are efficient in enforcing content-based security policies at a low computational cost.