Non-anonymized privacy preserving global and local anomaly detection in distributed systems
One example method may be performed in a distributed environment that includes a group of nodes configured to communicate with each other, and to communicate with a central node of the distributed environment. The example method includes performing, by two or more of the nodes, the operations of generating a public/private key pair, generating a secret shared key using the respective public keys of the other nodes, using the secret shared key to generate isometric transformations of data of the node, and transmitting the transformed data to the central node.
1 . A method, comprising:
in a distributed environment that includes a group of nodes configured to communicate with each other, and to communicate with a central node of the distributed environment, performing, by two or more of the nodes, operations comprising:
generating a public/private key pair;
generating a secret shared key using the respective public keys of the other nodes, and the secret shared key is unknown to the central node;
using the secret shared key as a seed of a random generator to generate transformation matrices that correspond to respective isometric transformations of node data;
wherein the transformation matrices are multiplied together to create a complete transformation matrix common to all the nodes, but unknown to the central node;
using the complete transformation matrix to transform the node data into transformed data; and
transmitting the transformed data, but not the node data, to the central node, and the transformed data is usable by the central node to identify an anomaly in the node data when the central node applies a time series discord finding algorithm to the transformed data.
2 . The method as recited in claim 1 , wherein the transformed data is non-anonymized but a privacy of the transformed data is preserved so that the data from which the transformed data was generated is not accessible by the central node.
3 . The method as recited in claim 1 , wherein the anomaly is a local anomaly, or a global anomaly.
4 . The method as recited in claim 1 , wherein the data comprises a multivariate or univariate time series.
5 . The method as recited in claim 1 , wherein the secret shared key is generated using a key agreement comprising a tuple of algorithms.
6 . The method as recited in claim 4 , wherein the tuple of algorithms comprises: KA.param, KA.gen, and KA.agree.
7 . The method as recited in claim 1 , wherein the isometric transformations comprise a reflection value, a rotation value, and a translation value.
8 . The method as recited in claim 1 , wherein the transformed data is indexed with information identifying the node that sent the transformed data.
9 . The method as recited in claim 1 , wherein the secret shared key is not known to the central node.
10 . In a distributed environment including two or more nodes configured to communicate with a central node of the distributed environment, a non-transitory storage medium having stored therein instructions that are executable by one or more hardware processors to perform operations comprising:
generating a public/private key pair;
generating a secret shared key using the respective public keys of the other nodes, and the secret shared key is unknown to the central node;
using the secret shared key as a seed of a random generator to generate transformation matrices that correspond to respective isometric transformations of node data;
wherein the transformation matrices are multiplied together to create a complete transformation matrix common to all the nodes, but unknown to the central node;
using the complete transformation matrix to transform the node data into transformed data; and
transmitting the transformed data, but not the node data, to the central node, and the transformed data is usable by the central node to identify an anomaly in the node data when the central node applies a time series discord finding algorithm to the transformed data.
11 . The non-transitory storage medium as recited in claim 10 , wherein the transformed data is non-anonymized but a privacy of the transformed data is preserved so that the data from which the transformed data was generated is not accessible by the central node.
12 . The non-transitory storage medium as recited in claim 10 , wherein the anomaly is a local anomaly, or a global anomaly.
13 . The non-transitory storage medium as recited in claim 10 , wherein the data comprises a multivariate or univariate time series.
14 . The non-transitory storage medium as recited in claim 10 , wherein the secret shared key is generated using a key agreement comprising a tuple of algorithms.
15 . The non-transitory storage medium as recited in claim 13 , wherein the tuple of algorithms comprises: KA.param, KA.gen, and KA.agree.
16 . The non-transitory storage medium as recited in claim 10 , wherein the isometric transformations comprise a reflection value, a rotation value, and a translation value.
17 . The non-transitory storage medium as recited in claim 10 , wherein the transformed data is indexed with information identifying the node that sent the transformed data.
18 . The non-transitory storage medium as recited in claim 10 , wherein the secret shared key is not known to the central node.