com.facebook.airlift.stats.cardinality

Class PrivateLpcaSketch

java.lang.Object

com.facebook.airlift.stats.cardinality.PrivateLpcaSketch

@NotThreadSafe
public class PrivateLpcaSketch
extends Object

A privacy-preserving cardinality sketch similar in spirit to HyperLogLog. This is essentially a data sketch that has buckets similar to those of HLL, but converted to binary values (above/below) relative to some threshold. This serves as a sort of data minimization. Additional noise is added to the sketch by randomly perturbing the buckets via randomized response. The final cardinality estimate is similar to the classical linear probabilistic counting algorithm (LPCA).

The final sketch (and the resulting cardinality estimate) is epsilon-differentially private, where epsilon = epsilonThreshold + epsilonRandomizedResponse.

Constructor Summary

Constructors

Constructor and Description
PrivateLpcaSketch(HyperLogLog hll, double epsilonThreshold, double epsilonRandomizedResponse)
PrivateLpcaSketch(HyperLogLog hll, double epsilonThreshold, double epsilonRandomizedResponse, RandomizationStrategy randomizationStrategy)
PrivateLpcaSketch(io.airlift.slice.Slice serialized)
PrivateLpcaSketch(io.airlift.slice.Slice serialized, RandomizationStrategy randomizationStrategy)

Method Summary

Modifier and Type	Method and Description
long	cardinality()
int	estimatedSerializedSize()
int	getNumberOfBuckets()
int	getThreshold()
io.airlift.slice.Slice	serialize()
void	update(HyperLogLog hllOther) Updates current sketch by adding data from a second HyperLogLog

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

PrivateLpcaSketch

public PrivateLpcaSketch(HyperLogLog hll,
                         double epsilonThreshold,
                         double epsilonRandomizedResponse)

PrivateLpcaSketch

public PrivateLpcaSketch(io.airlift.slice.Slice serialized)

PrivateLpcaSketch

public PrivateLpcaSketch(HyperLogLog hll,
                         double epsilonThreshold,
                         double epsilonRandomizedResponse,
                         RandomizationStrategy randomizationStrategy)

PrivateLpcaSketch

public PrivateLpcaSketch(io.airlift.slice.Slice serialized,
                         RandomizationStrategy randomizationStrategy)

Method Detail

cardinality

public long cardinality()

estimatedSerializedSize

public int estimatedSerializedSize()

getNumberOfBuckets

public int getNumberOfBuckets()

getThreshold

public int getThreshold()

serialize

public io.airlift.slice.Slice serialize()

update

public void update(HyperLogLog hllOther)

Updates current sketch by adding data from a second HyperLogLog

Parameters:

hllOther - HyperLogLog of data to add to current sketch