Session 2.2 - Doctoral Symposium: Presentations by Ph.D. Students


Chair: Dr. Michael Rodeh, VP, Infrastructure Technologies, IBM Corporate


Lecture Title: A Cultural Sensitive Agent for Human-Computer Negotiation


Speaker: Galit Haim
, Bar-Ilan University.


Authors: Galit Haim, Bar-Ilan University ; Ya’akov (Kobi) Gal, en-Gurion University of the Negev, Israel ; Michele Gelfand, University of Maryland, USA ; Sarit Kraus, Bar Ilan University, Israel.

 

Abstract: People's cultural background has been shown to affect the way they reach agreements in negotiation and how they fulfill these agreements. This paper presents a novel agent de- sign for negotiating with people from different cultures. Our setting involved an alternating-offer protocol that allowed parties to choose the extent to which they kept each of their agreements during the negotiation. A challenge to designing agents for such setting is to predict how people reciprocate their actions over time despite the scarcity of prior data of their behavior across different cultures. Our methodology addresses this challenge by combining a decision theoretic model with classical machine learning techniques to predict how people respond to offers, and the extent to which they fulfill agreements. The agent was evaluated empirically by playing with 157 people in three countries- Lebanon, the U.S., and Israel - in which people are known to vary widely in their negotiation behavior. The agent was able to out- perform people in all countries under conditions that varied how parties depended on each other at the onset of the negotiation. This is the first work to show that a computer agent can learn to outperform people when negotiating in three countries representing different cultures.

 


Lecture Title:  Deterring attacks against Critical IT Infrastructure


Speaker: Ofer Hermoni
, Ben Gurion University of the Negev.

 

Abstract: Cyber-attacks and many types of digital fraud are almost always risk-free. Whether targeting government services, _nancial institutions or corporate assets, an attacker can sit in the comfort and safety of his home and mount one attack after the other. Protected from identi_cation by the virtual anonymity of the Internet and from legal proceedings by being in a di_erent jurisdiction or country from the target, the greatest risk for most attackers is that their attack may fail. The typical approach to mitigating attacks is identi_cation and prevention. Operators of IT infrastructure employ various technological and social means to ensure that the tra_c they receive is not malicious. If an attack is detected then it is blocked. This approach is analogous to locked doors, alarm system and security cameras in the physical world. However, in the real world, identi_cation and prevention are complemented by deterrence. A criminal runs a real risk of being caught by the police and then prosecuted and penalized by the legal system. In this abstract we propose a new paradigm for deterring remote IT attackers. The basic idea is to add a dimension of liability to the interaction between a client and an IT service. The client and server reach an agreement that regulates the client behavior, e.g. by demanding no cyber-attacks or fraud attempts against the server. The client then posts a signed and encrypted digital bond with the server. As long as the client abides by the agreement, the server can't decrypt the bond and receive the associated payment. If the client violates the agreement then the server obtains the information necessary to decrypt the bond. The server is unable to impersonate the client and decrypt the digital bond by forging a cyber-attack or a fraud attempt, because the client's messages are signed by the client's private-key, which is not known to the server. Conditional Anonymity is one variant of our generic scheme. We use the notion of arbitrators in a Peer-to-Peer (P2P) network to enforce the client-server agreement. Arbitrators are P2P semi- trusted entities that function as a jury in the technology court of law. The communicating parties, users and servers, agree in the initial phase on a set of arbitrators that they trust (reputation systems may support their choice). Then, the user divides its identity into shares and sends each share to one arbitrator, such that only a large enough number of arbitrators can reveal the identity of the user. The communication between the user and the server is performed in an undeniable manner, which means that the server can convince the arbitrators that the user misbehaved. In case the server _nds a violation of the terms of the policy, the server proves to the arbitrators that a violation took place and the arbitrators reconstruct the user's identity. The conditional nature of our scheme makes it attractive for legitimate users. They do not need to trust the server. The commitment is executed and the users' digital goods are lost only if the users actively violate a signed agreement, which is not the aim of the typical law abiding user.

 

*Joint work with Shlomi Dolev and Niv Gilboa

 


Lecture Title:  OPM-based Model Verification Framework with Application to Molecular Biology


Speaker: Yudit Somekh
, Technion.

 

Abstract: A myriad of detailed pieces of knowledge regarding the structure and function of the living cell have been accumulating at an alarmingly increasing rate. Emphasis is shifting from the study of a single molecular process to cellular pathways, cycles, and the entire cell as a system. A framework for supporting the biological researcher for hypotheses verification is proposed. The framework includes molecular biological systems modeling and verification against pertinent literature. Object-Process Methodology (OPM) is a holistic graphical modeling methodology that combines the behavioral and structural aspects of a system in a single model. The OPM methodology includes OPM-based development process, OPM Case Tool (OPCAT), and a modeling language. In this work, we propose an OPM-based verification framework for molecular biology systems. The framework includes a set of translation rules from OPM to a finite-state transition system and classification of mechanistic requirements derived from biological experimental findings. The framework is exemplified on the gene expression system.