You have come to an old website. Our new URL is ths.rwth-aachen.de. Please update your bookmarks!




Accelerating Parametric Probabilistic Verification

This is a webpage for our project concerning parametric probabilistic verification.  If you encounter any problems or if you have any questions, please contact Nils Jansen.

 

Documentation

We have submitted a paper about model checking parametric discrete-time Markov chains to QEST 2014. A technical report including all proofs is available here.

 

Our first work in this project was done within the bachelor thesis of Matthias Volk, which is available here.

 

Implementation

The sources of our implementation can be found here. Installation instructions can be found in the included README file.

 

A linux binary is available here.

 

Involved Persons

The people involved with this project are Nils JansenFlorian Corzilius, Matthias Volk, Erika ÁbrahámRalf WimmerJoost-Pieter Katoen and Bernd Becker

 

Benchmarks

We give three different sets of benchmarks. Short descriptions and downloads are listed below, more benchmarks for PDTMCs are available at the homepage of PARAM.

 

The Bounded retransmission protocol [HSV94] sends a file by dividing it into a number of N chunks. Only a bounded number MAX of retransmissions for each chunk is allowed. Additionally, there are two lossy channels K and L with reliability pK and pL for sending data and acknowledgments respectively. The original version of this benchmark is available at the PRISM homepage. A parametric version is available at the PARAM homepage, where pK and pL are used as parameters in the model. We compute the probability that the sender successfully transmits a chunk but does not report that. The set of instances we tested for the paper is available here.

 

The Crowds protocol [RR98] is designed for anonymous communication in the internet via random routing. Sending a message to a specific user allows for two different actions. The message either is sent directly to this user or first is sent to a random user, who acts a router by only forwarding the message. Therefore, an intruder can not be certain if a message was sent by a specific user or if the sender only forwarded the massage. Thus anonymity can be preserved. In our test cases we have N honest users, M dishonest users and therefore a percentage of B=M/(M+N) untrustworthy users. R many messages are sent during the whole process. The probability of sending the message directly to its destination is 1-pf, whereas the message is forwarded with probability pf. We use pf and B as parameters in our model and are interested in the probability, that one user is identified more often by a bad member than every other. The original version of this benchmark is available at the PRISM homepage. A parametric version is available at the PARAM homepage. The set of instances we tested for the paper is available here.

 

NAND multiplexing [HJ02] deals with constructing reliable computation from unreliable hardware. For performing a computation, a NAND unit is multiplexed by copying it N times with own input and output. Next, K restorative stages are introduced to reduce the degradation after the NAND computation. In our model we have the parameter perr as the probability for NAND not working correctly and pinp as the probability for the initial inputs to be stimulated. We are now interested in the probability that less than 10% of the computed outputs are erroneous. The original version of this benchmark is available at the PRISM homepage. The set of instances we tested for the paper is available here.

 

[HSV94] Helmink, L., Sellink, M., Vaandrager, F.: Proof-checking a data link protocol. In: Proc. of TYPES. Volume 806 of LNCS, Springer (1994) 127–165 

 

[RR98] Reiter, M.K., Rubin, A.D.: Crowds: Anonymity for web transactions. ACM Trans. on Information and System Security 1(1) (1998) 66–92

 

[HJ02] Han, J., Jonker, P.: A system architecture solution for unreliable nanoelectronic devices. IEEE Transactions on Nanotechnology (2002) 201–208 


Publications related to Parametric Probabilistic Verification

2013
DownloadNils Jansen, Florian Corzilius, Matthias Volk, Ralf Wimmer, Erika Abraham, Joost-Pieter Katoen, Bernd Becker. Accelerating Parametric Probabilistic Verification. Technical report at Cornell University number arXiv:1312.3979, 2013.