Paper notes (1)--Where's Wally? Precise User Discovery Attacks on location Proximity Services

Abstract:

Location-Close service The widespread use of social and mobile networks is based on the balance of usability and user privacy, but it raises the risk of triangulation attacks. This article systematically discusses the prevention of such attacks, including the formalization of the problem under different neighboring models, effective attacks against different models, and the exact number of queries required for the attack, and experiments for practical applications.

A) Modeling the attack: UDP, known as the Euclidean plane area A with point P and a black box providing neighborhood information, find the position of point P

Neighborhood (proximity Oracle) Definition: An area centered on a point

The original problem is two-part:

1) Disk coverage: A with a minimum R-neighborhood coverage

The minimum dominant set (MDS) problem on the UDG (Unit Disk Graph) is Np-hard, but there is a linear time 5-Approximate random algorithm (the result is not more than five times times the optimal solution gap)

Approximate algorithm: Random fetch points to the dominant set, remove all adjacent points, repeat to the graph is empty. The complexity of Time is O (| v|)

UDG: There are many sampling points on the plane, if the distance between two points is less than R there is an edge, so that the minimum dominating set will be able to cover all the sampling points with R-Neighborhood

For Max-coverage, the distance between points in the dominating set are at least

2) Disk Search: Find out which neighborhood p is in

The points in each disk can be completely covered by an "add-in" rectangle, with a binary algorithm that can be resolved at O (rlogr) time (the number of queries is LOGR)

So the total number of queries is

II) RUDP (rounding User Discovery problem)

For different distances of P and p_u, social networks usually return different distance values rather than fixed r, thus studying rounding Class family to solve this problem

RCF consists of a series of tuple, for rounding Value,i1,..., in constitutes a partition of r+, and

By continuous triangulation, the range of the next point is narrowed down to r=delta_1, thus the total elapsed time of the original algorithm is (| S| is the size of rounding class family, which is clearly the number of queries)

III) randomized User Discovery problem

The results of a query on a point are subject to a random distribution (the result of each return contains Gaussian noise), and after some mathematical processing, the error of solving the RANDUDP problem is

(iv) Practical issues

Query space: Large, reduce query space through personal information

Contact: Forged identity plus friend attack

Attack detection: This type of service has a mechanism for detecting forged locations, using a camouflage mechanism (see the man who is there:validating check-ins in location-based services)

Accuracy: Related to GPS accuracy

Projection error: Coordinates need to be obtained using a suitable projection method, with equidistant conic projection (equidistant conic projection)

Paper notes (1)--Where's Wally? Precise User Discovery Attacks on location Proximity Services