Efficient detection of network anomalies
View Patent ↗Techniques of identifying anomalous behavior on an electronic network involve iteratively combining groups of adjacent bins of a histogram in such a way as to minimize a measure of error in the histogram. Along these lines, a user behavior analytics server represents a user behavior factor with a histogram. The UBA server reduces a number of bins in the histogram by iteratively selecting groups of adjacent bins for combination. Upon each iteration, the group of bins that is selected for combination is the group which, when its bins are combined, minimizes differences between the values of the bins in that group and a value of the combined bin.
1. A method of identifying anomalous use of an electronic network, the method comprising:
generating, by processing circuitry, a histogram from values of a user behavior factor of a user, the histogram including bins, each bin (i) representing a range of the user behavior factor and (ii) having a value indicative of a relative frequency of the behavior factor in the range represented by that bin;
receiving, by the processing circuitry after generating the histogram, network data indicating that an alleged user is interacting with the electronic network; and
in response to receipt of the network data, (i) identifying, by the processing circuitry, a value of a behavior factor from the network data, (ii) comparing the value of the behavior factor with the histogram to identify a bin of the histogram that encompasses the behavioral factor; and (iii) indicating anomalous use of the electronic network in response to the identified bin of the histogram having a lower value than values of other bins in the histogram,
wherein generating the histogram includes performing multiple bin-reduction iterations, each bin reduction iteration including (i) selecting, by the processing circuitry, a group of adjacent bins of the histogram for combination into a combined bin, the group of bins selected to minimize differences between a value of the combined bin and the value of each of the bins in that group, and (ii) combining the bins of the selected group to form the combined bin, the combined bin replacing the bins of the selected group in the histogram and the value of the combined bin being based on the values of the bins in the selected group, and
wherein the network data indicating that an alleged user is interacting with the electronic network includes network data indicating that the user is initiating a VPN (Virtual Private Network) session to obtain remote access to the electronic network.
2. A method as in claim 1 , further comprising, prior to generating the histogram, obtaining prior network data from the electronic network, the prior network data including multiple values of the user behavior factor of the user.
3. A method as in claim 1 , wherein selecting a group of adjacent bins of the histogram includes:
for each group of N adjacent bins of the histogram, (i) performing a trial combination operation on that group to produce a trial combined bin and (ii) computing an error based on a difference between the value of each bin of that group and a value of the trial combined bin; and
providing the selected group as the group of N adjacent bins for which the computed error is a minimum over all the groups of N adjacent bins of the histogram.
4. A method as in claim 3 , wherein each bin of the histogram also has a width equal to a difference between a maximum and a minimum of the range of the user behavior factor represented by that bin,
wherein a width of each trial combined bin is a sum of the widths of the adjacent bins in that group, and
wherein the value of the trial combined bin is a weighted average of the values of the adjacent bins in that group.
5. A method as in claim 3 , wherein computing the error includes, for each group of adjacent bins of the histogram:
generating a respective absolute difference between the value of each bin in that group and the value of that trial combined bin; and
producing, as the error, a weighted average of the generated differences in that group.
6. A method as in claim 5 , further comprising terminating further bin-reduction iterations in response to a minimum error produced across all groups of adjacent bins in a current bin-reduction iteration being greater than a predetermined threshold.
7. A method as in claim 5 , further comprising terminating further bin-reduction iterations in response to a number of bin-reduction iterations reaching a predetermined maximum number of bin-reduction iterations.
8. A method as in claim 1 , wherein N equals two, such that each group of N adjacent bins is a pair of adjacent bins.
9. A computer program product including a non-transitory, computer-readable storage medium which stores executable code, which when executed by a computer, causes the computer to perform a method of identifying anomalous use of an electronic network, the method comprising:
generating a histogram from values of a user behavior factor of a user, the histogram including bins, each bin (i) representing a range of the user behavior factor and (ii) having a value indicative of a relative frequency of the behavior factor in the range represented by that bin;
receiving, after generating the histogram, network data indicating that an alleged user is interacting with the electronic network; and
in response to receipt of the network data, (i) identifying a value of a behavior factor from the network data, (ii) comparing the value of the behavior factor with the histogram to identify a bin of the histogram that encompasses the behavioral factor; and (iii) indicating anomalous use of the electronic network in response to the identified bin of the histogram having a lower value than values of other bins in the histogram,
wherein generating the histogram includes performing multiple bin-reduction iterations, each bin reduction iteration including (i) selecting a group of adjacent bins of the histogram for combination into a combined bin, the group of bins selected to minimize differences between a value of the combined bin and the value of each of the bins in that group, and (ii) combining the bins of the selected group to form the combined bin, the combined bin replacing the bins of the selected group in the histogram and the value of the combined bin being based on the values of the bins in the selected group, and
wherein indicating anomalous use of the electronic network includes sending an alert to a UBA (User Behavior Analytics) administrator.
10. A computer program product as in claim 9 , wherein the method further comprises, prior to generating the histogram, obtaining prior network data from the electronic network, the prior network data including multiple values of the user behavior factor of the user.
11. A computer program product as in claim 9 , wherein selecting a group of adjacent bins of the histogram includes:
for each group of N adjacent bins of the histogram, (i) performing a trial combination operation on that group to produce a trial combined bin and (ii) computing an error based on a difference between the value of each bin of that group and a value of the trial combined bin; and
providing the selected group as the group of N adjacent bins for which the computed error is a minimum over all the groups of N adjacent bins of the histogram.
12. A computer program product as in claim 11 , wherein each bin of the histogram also has a width equal to a difference between a maximum and a minimum of the range of the user behavior factor represented by that bin,
wherein a width of each trial combined bin is a sum of the widths of the adjacent bins in that group, and
wherein the value of the trial combined bin is a weighted average of the values of the adjacent bins in that group.
13. A computer program product as in claim 11 , wherein computing the error includes, for each group of adjacent bins of the histogram:
generating a respective absolute difference between the value of each bin in that group and the value of that trial combined bin; and
producing, as the error, a weighted average of the generated differences in that group.
14. A computer program product as in claim 13 , wherein the method further comprises terminating further bin-reduction iterations in response to a minimum error produced across all groups of adjacent bins in a current bin-reduction iteration being greater than a predetermined threshold.
15. A computer program product as in claim 13 , further comprising terminating further bin-reduction iterations in response to a number of bin-reduction iterations reaching a predetermined maximum number of bin-reduction iterations.
16. A computer program product as in claim 9 , wherein N equals two, such that each group of N adjacent bins is a pair of adjacent bins.
17. An electronic apparatus constructed and arranged to identify anomalous use of an electronic network, the electronic apparatus including:
a network interface;
memory; and
controlling circuitry couple to the memory, the controlling circuitry being constructed and arranged to:
generate a histogram from values of a user behavior factor of a user, the histogram including bins, each bin (i) representing a range of the user behavior factor and (ii) having a value indicative of a relative frequency of the behavior factor in the range represented by that bin;
receive, after generating the histogram, network data indicating that an alleged user is interacting with the electronic network; and
in response to receipt of the network data, (i) identify a value of a behavior factor from the network data, (ii) compare the value of the behavior factor with the histogram to identify a bin of the histogram that encompasses the behavioral factor; and (iii) indicate anomalous use of the electronic network in response to the identified bin of the histogram having a lower value than values of other bins in the histogram,
wherein the controlling circuitry constructed and arranged to generate the histogram is further constructed and arranged to perform multiple bin-reduction iterations, each bin reduction iteration including (i) selecting, by the processing circuitry, a group of adjacent bins of the histogram for combination into a combined bin, the group of bins selected to minimize differences between a value of the combined bin and the value of each of the bins in that group, and (ii) combining the bins of the selected group to form the combined bin, the combined bin replacing the bins of the selected group in the histogram and the value of the combined bin being based on the values of the bins in the selected group, and
wherein the network data indicating that an alleged user is interacting with the electronic network includes network data indicating that the user is initiating a VPN (Virtual Private Network) session to obtain remote access to the electronic network.
18. An electronic apparatus as in claim 17 , wherein the controlling circuitry is further constructed and arranged to, prior to generating the histogram, obtain prior network data from the electronic network, the prior network data including multiple values of the user behavior factor of the user.
19. An electronic apparatus as in claim 17 , wherein the controlling circuitry constructed and arranged to select the group of adjacent bins of the histogram is further constructed and arranged to:
for each group of N adjacent bins of the histogram, (i) perform a trial combination operation on that group to produce a trial combined bin and (ii) compute an error based on a difference between the value of each bin of that group and a value of the trial combined bin; and
provide the selected group as the group of N adjacent bins for which the computed error is a minimum over all the groups of N adjacent bins of the histogram.
20. An electronic apparatus as in claim 19 , wherein each bin of the histogram also has a width equal to a difference between a maximum and a minimum of the range of the user behavior factor represented by that bin,
wherein a width of each trial combined bin is a sum of the widths of the adjacent bins in that group, and
wherein the value of the trial combined bin is a weighted average of the values of the adjacent bins in that group.