CS 8803 Week 3 Paper Critque

Week 3 Critiques



1) Artificial Project 1 Uninformed and Informed Search Algorithms.

2) Artificial Intelligence Project on Alpha Beta Pruning.

3) Artificial Intelligence

4) Computer Security ( Bell LaPadula Model)

5) Critical Essay and Analysis Part 2

6) Critical Essay and Analysis Part 1.

7) CS8803 AIA @ Georgia Institute of Technology

8) Explicit and Implicit DLL Linking

9) Field Hiding and Method Overriding

10) Fun With Pointers in C++

11) Resume Builder

12) AIA Week 5


CS8803 Course Reading Summaries 
Title: Search and Replication in Unstructured
Peer-Peer Networks. 
- The paper seems to be the first at
time of publication in understanding the 
properties of 
scalable search algorithms, replication 
and network topologies for decentralized 
unstructured peer to peer network. 

- The paper puts forward three critical 
yet essential properties i.e. 1) adaptive 
termination, 2) minimizing message duplication 3) 
small granularity of coverage. Using these three 
properties as a basis, the algorithms of 
Expanding Ring and K-Walker are presented. 
This allows the user to evaluate all the different 
kinds of algorithms using the same standard set of 

- The paper presents simulation for
the new 
methodologies proposed by the authors in 
and replication. For example, in Section 
2.3 the 
flooding of the network in Gnutella does 
not satisfy
 any of the three above mentioned 
properties and 
does not scale well. The expansion ring 
approach is
 significantly better than flooding by 
using adaptive 
termination while K-Walker is the best 
as it reduced 
the granularity of coverage. These 
simulations are a 
fundamental strength of the paper, as 
the authors have 
not only presented the new algorithm 
methodology but
 proof (simulation) that they are significantly 
better in reducing flooding. 

-The paper only uses 4 topologies mentioned in 
section 2 of the paper with an average of only 
4000 nodes per topology. In reality on the actual 
Internet, there are P2P nodes of a higher magnitude 
with heterogeneous topologies where K-Walker search 
techniques may not be scalable. 

-The paper does not model the various search 
algorithms with certain network topologies and study
 them analytically. 

-It would be interesting if the paper had 
presented the tradeoff between the two metrics in a 
distributed system of Square Root replication 
(minimizes search size) and Proportional replication 
(optimizes balance load). 

- It is an excellent research for unstructured
 search and replication, but the paper must have 
presented statistics in comparing the three kinds 
of P2P i.e. Centralized, Decentralized (Structured)
 and Decentralized Unstructured in order to gauge
 whether in practice K-Walker search and the new 
replication methodology will make a significant 
difference if the number of people using this kind 
of P2P is relatively small. 

CS8803 Course Reading Summaries 
Title: TrustMe: Anonymous Management of Trust 
Relationship in Decentralized P2P Systems. 

-The paper attempts to answer two important 
questions i.e. 1) Where should the trust value 
be stored? 2) How to securely access other peer's 
trust value. Question 1 is challenging enough, 
because of the decentralized nature of the Peer-Peer
 network. The paper explicitly lays out the questions
 it will attempt to answer ensuring the following 
properties remain invariant a) Security b) Reliability 
c) Accountability. This clear and concise scope and
 objective of the paper allows the reader to understand 
the context of the material to follow in the paper. 

- The paper solves the problem of storing user 
ratings as mutual anonymity. If nobody knows who 
the trust store is (including the peer whose trust
 value is being stored), there is hardly a chance 
that the trust store will be attacked.

-The paper highlights some of the shortcomings 
of Eigenrep such as Insecure Communication, DHT 
Threats, No Anonymity and Group Threats and at the 
same time TrustMe is resistant to these threats as
 it employs a random assignment of Trust Holding 
Agent peers and uses Public Key mechanism to prevent
 loss of anonymity. This reasoning allows the user
 to understand how TrustMe fairs and ranks with its 

-There are various challenging issues that need to
 handled e.g. how to select good trust stores, 
how to access ratings from them without knowing
 who/where they are? TrustMe provides all the 

-Section 3.3.2 presents some of the common attack 
scenario such as �Manipulating Reply Message�,
�Manipulating Proof of Interaction Messages� and how
 TrustMe maintains security, reliability and 

-One apparent weakness of the paper, is that in
 the user based approach a potential malevolent 
user can harmfully/intentionally harm a user 
providing good (QOS) by giving it low ratings. OR a
 group of malicious users together harm a good QOS
 provider. How will TrustMe address this?

-The trust value should be only visible for a user
 after it has received certain number of trusts. 
This is because, if a user receives a bad/low rating
 out of 1 users, then it has a 100% low rating. 
The paper should address this and use some sort of 
statistical/variance method to delete off votes that
 have large variance. 

-As security is cornerstone of TrustMe, 
the paper should have explored various other 
cryptographic methodologies that can be more 
efficient as compared to Public Key cryptography.