25-08-2017, 09:32 PM
On Privacy of Encrypted Speech Communications
Privacy of Encrypted.pptx (Size: 198.11 KB / Downloads: 24)
INTRODUCTION
SPEECH COMMUNICATION HAVE INCREASED POPULARITY
USED IN TELEPHONE NETWORKS
BASED ON ENCRYPTED SPEECH COMMUNICATION
COMMUNICATION USES INFORMATION AS PACKETS
THE GOAL OF THE PROJECT IS TO FIND THE SPEAKERS OF THE ENCRYPTED SPEECH COMMUNICATION
EXISTING SYSTEM
BY PROVIDING A CLASS OF PASSIVE TRAFFIC ANALYSIS ATTACK WE CAN COMPROMISE THIS COMMUNICATION
CAN DETECT SPEAKERS OF COMMUNICATION
TO PROTECT CONFIDENTIALITY ,TOOLS ARE :
Zfone: To encrypt voice packets
SRTP: Secure version of Real Time Transport Protocol
ANOTHER IDEA IS TO CREATE ANONYMITY NETWORKS FOR ANONYMOUS SPEECH COMMNCN, LIKE
Tor , JAP
MECHANISM FOR CREATE ANONYMITY NETWORK:REROUTING PACKETS THROUGH LONGEST PATH
PROPOSED SYSTEM
Propose a new class of traffic analysis attacks based on
packet timing information
and it can detect speakers of speech communications made with different codecs.
It experiments over different type of networks
commercial anonymity networks
campus networks.
Speech coding
An analog voice signal is first converted into a voice data stream by a chosen codec
compression is used to reduce the data rate.
It is packetized in small units and encapsulated in a packet
We use CBR codecs
Speaker Detection:
Input:
Alice’s HMM
the feature vectors generated
Output : intermediate detection result.
The detection step can be divided into two phases:
1) First, the likelihood of each feature vector is calculated with the trained HMM.
2) The trace with the highest likelihood is declared as Alice’s trace if the intersection step is not used. To improve the detection accuracy, the intermediate detection results can be fed into the optional intersection attack step.
Cross – Codec Detection
The training traces and the traces to be detected are generated with different codecs.
It is important because:
1) Practically training traces and the traces to be detected can be collected from speech communications made with different codecs.
2) Since speech packets are encrypted and possibly padded to a fixed length