Identifying and protecting against computer security threats while preserving privacy of individual client devices using condensed local differential privacy (CLDP)
Identifying and protecting against computer security threats while preserving privacy of individual client devices using condensed local differential privacy (CLDP). In one embodiment, a method may include mapping non-ordinal data values to ordinal data values, generating a first ordering scheme for the ordinal data values, accessing actual non-ordinal data values, converting the actual non-ordinal data values to actual ordinal data values according to the mapping, generating first perturbed ordinal data values by adding noise, and aggregating the first perturbed ordinal data values. The method may also include generating a second ordering scheme for the ordinal data values by denoising and refining the aggregated first perturbed ordinal data values, generating second perturbed ordinal data values by adding noise, aggregating the second perturbed ordinal data values, analyzing the aggregated second perturbed ordinal data values to identify a computer security threat, and in response, thwarting the computer security threat.
1. A computer-implemented method for identifying and protecting against computer security threats while preserving privacy of individual client devices using condensed local differential privacy (CLDP), at least a portion of the method being performed by a computer device comprising at least one processor, the method comprising:
mapping, at a remote server device, non-ordinal data values to ordinal data values;
generating, at the remote server device, a first ordering scheme for the ordinal data values;
sending, to local client devices, the mapping and the first ordering scheme;
accessing, at each of the local client devices, an actual non-ordinal data value;
converting, at each of the local client devices, the actual non-ordinal data value to an actual ordinal data value according to the mapping;
generating, at each of the local client devices, a first perturbed ordinal data value by adding noise to the actual ordinal data value, an amount of noise being added, according to the first ordering scheme, to each actual data value being probabilistically computed such that a probability of noise being added decreases as an amount of added noise increases;
sending, from each of the local client devices, the first perturbed ordinal data value to the remote server device;
aggregating, at the remote server device, the first perturbed ordinal data values to at least partially cancel out aggregate noise of the aggregated first perturbed ordinal data values at a population level;
generating, at the remote server device, a second ordering scheme for the ordinal data values by denoising and refining the aggregated first perturbed ordinal data values;
sending, to local client devices, the second ordering scheme;
generating, at each of the local client devices, a second perturbed ordinal data value by adding noise to the actual ordinal data value, an amount of noise being added, according to the second ordering scheme, to each actual data value being probabilistically computed such that a probability of noise being added decreases as an amount of added noise increases;
sending, from each of the local client devices, the second perturbed ordinal data value to the remote server device;
aggregating, at the remote server device, the second perturbed ordinal data values to at least partially cancel out aggregate noise of the aggregated second perturbed ordinal data values at a population level;
analyzing, at the remote server device, using CLDP, the aggregated second perturbed ordinal data values to identify a computer security threat to the local client devices while preserving privacy of the actual data values by preventing the remote server device from inferring the actual data value of any of the local client devices; and
in response to identifying the computer security threat, protecting against the computer security threat by performing, at one or more of the local client devices, a remedial action to protect the one or more local client devices from the computer security threat.
2. The method of claim 1 , wherein the performing, at the one or more local client devices, of the remedial action comprises one or more of blocking the one or more local client devices from accessing a network, rolling back one or more changes at the one or more local client devices that were made in response to the computer security threat, or temporarily freezing activity at the one or more local client devices, or some combination thereof.
3. The method of claim 1 , wherein the noise is added to each actual ordinal data value with a probability of noise being added decreasing exponentially as an amount of added noise increases.
4. The method of claim 1 , wherein the computer security threat comprises a malware outbreak.
5. The method of claim 4 , wherein:
the actual non-ordinal data values comprise names of operating systems being executed on the local client devices; and
the adding of the noise to the actual ordinal data values prevents the remote server device from inferring the name of the operating system being executed on any of the local client devices.
6. The method of claim 4 , wherein the analyzing comprises analyzing, at the remote server device, using CLDP, the aggregated second perturbed ordinal data values to identify a timing of a malware infecting the local client devices.
7. The method of claim 1 , further comprising:
accessing, at each of the local client devices, a sequence of actual non-ordinal data values;
generating, at each of the local client devices, a sequence of perturbed non-ordinal data values by adding noise to one or more of the actual data values in the sequence of actual non-ordinal data values, an amount of noise being added, according to the mapping and the second ordering scheme, to each actual data value being probabilistically computed such that a probability of noise being added decreases as an amount of added noise increases
sending, from each of the local client devices, the sequence of perturbed non-ordinal data values to the remote server device;
aggregating, at the remote server device, the sequences of perturbed non-ordinal data values to at least partially cancel out aggregate noise of the aggregated sequences of perturbed non-ordinal data values at a population level;
analyzing, at the remote server device, using CLDP, the aggregated sequences of perturbed non-ordinal data values to identify a second computer security threat to the local client devices while preserving privacy of the sequences of actual non-ordinal data values by preventing the remote server device from inferring the sequence of actual non-ordinal data values, and an actual length of the sequence of actual non-ordinal data values, of any of the local client devices; and
in response to identifying the second computer security threat, protecting against the second computer security threat by performing, at one or more of the local client devices, a second remedial action to protect the one or more local client devices from the second computer security threat.
8. One or more non-transitory computer-readable media comprising one or more computer-readable instructions that, when executed by one or more processors of a computer device, cause the computer device to perform a method for identifying and protecting against computer security threats while preserving privacy of individual client devices using condensed local differential privacy (CLDP), the method comprising:
mapping, at a remote server device, non-ordinal data values to ordinal data values;
generating, at the remote server device, a first ordering scheme for the ordinal data values;
sending, to local client devices, the mapping and the first ordering scheme;
accessing, at each of the local client devices, an actual non-ordinal data value;
converting, at each of the local client devices, the actual non-ordinal data value to an actual ordinal data value according to the mapping;
generating, at each of the local client devices, a first perturbed ordinal data value by adding noise to the actual ordinal data value, an amount of noise being added, according to the first ordering scheme, to each actual data value being probabilistically computed such that a probability of noise being added decreases as an amount of added noise increases;
sending, from each of the local client devices, the first perturbed ordinal data value to the remote server device;
aggregating, at the remote server device, the first perturbed ordinal data values to at least partially cancel out aggregate noise of the aggregated first perturbed ordinal data values at a population level;
generating, at the remote server device, a second ordering scheme for the ordinal data values by denoising and refining the aggregated first perturbed ordinal data values;
sending, to local client devices, the second ordering scheme;
generating, at each of the local client devices, a second perturbed ordinal data value by adding noise to the actual ordinal data value, an amount of noise being added, according to the second ordering scheme, to each actual data value being probabilistically computed such that a probability of noise being added decreases as an amount of added noise increases;
sending, from each of the local client devices, the second perturbed ordinal data value to the remote server device;
aggregating, at the remote server device, the second perturbed ordinal data values to at least partially cancel out aggregate noise of the aggregated second perturbed ordinal data values at a population level;
analyzing, at the remote server device, using CLDP, the aggregated second perturbed ordinal data values to identify a computer security threat to the local client devices while preserving privacy of the actual data values by preventing the remote server device from inferring the actual data value of any of the local client devices; and
in response to identifying the computer security threat, protecting against the computer security threat by performing, at one or more of the local client devices, a remedial action to protect the one or more local client devices from the computer security threat.
9. The one or more non-transitory computer-readable media of claim 8 , wherein the performing, at the one or more local client devices, of the remedial action comprises one or more of blocking the one or more local client devices from accessing a network, rolling back one or more changes at the one or more local client devices that were made in response to the computer security threat, or temporarily freezing activity at the one or more local client devices, or some combination thereof.
10. The one or more non-transitory computer-readable media of claim 8 , wherein the noise is added to each actual ordinal data value with a probability of noise being added decreasing exponentially as an amount of added noise increases.
11. The one or more non-transitory computer-readable media of claim 8 , wherein the computer security threat comprises a malware outbreak.
12. The one or more non-transitory computer-readable media of claim 11 , wherein:
the actual non-ordinal data values comprise names of operating systems being executed on the local client devices; and
the adding of the noise to the actual ordinal data values prevents the remote server device from inferring the name of the operating system being executed on any of the local client devices.
13. The one or more non-transitory computer-readable media of claim 11 , wherein the analyzing comprises analyzing, at the remote server device, using CLDP, the aggregated second perturbed ordinal data values to identify a timing of a malware infecting the local client devices.
14. The one or more non-transitory computer-readable media of claim 8 , wherein the method further comprises:
accessing, at each of the local client devices, a sequence of actual non-ordinal data values;
generating, at each of the local client devices, a sequence of perturbed non-ordinal data values by adding noise to one or more of the actual data values in the sequence of actual non-ordinal data values, an amount of noise being added, according to the mapping and the second ordering scheme, to each actual data value being probabilistically computed such that a probability of noise being added decreases as an amount of added noise increases
sending, from each of the local client devices, the sequence of perturbed non-ordinal data values to the remote server device;
aggregating, at the remote server device, the sequences of perturbed non-ordinal data values to at least partially cancel out aggregate noise of the aggregated sequences of perturbed non-ordinal data values at a population level;
analyzing, at the remote server device, using CLDP, the aggregated sequences of perturbed non-ordinal data values to identify a second computer security threat to the local client devices while preserving privacy of the sequences of actual non-ordinal data values by preventing the remote server device from inferring the sequence of actual non-ordinal data values, and an actual length of the sequence of actual non-ordinal data values, of any of the local client devices; and
in response to identifying the second computer security threat, protecting against the second computer security threat by performing, at one or more of the local client devices, a second remedial action to protect the one or more local client devices from the second computer security threat.
15. A remote server device comprising:
one or more processors; and
one or more non-transitory computer-readable media comprising one or more computer-readable instructions that, when executed by the one or more processors, cause the remote server device to perform a method for identifying and protecting against computer security threats while preserving privacy of individual client devices using condensed local differential privacy (CLDP), the method comprising:
mapping, at the remote server device, non-ordinal data values to ordinal data values;
generating, at the remote server device, a first ordering scheme for the ordinal data values;
sending, to local client devices, the mapping and the first ordering scheme;
receiving, at the remote server device from local client devices, first perturbed ordinal data values, the first perturbed ordinal data values resulting from actual non-ordinal data values being accessed, the actual non-ordinal data values being converted to actual ordinal data values according to the mapping, and the first perturbed ordinal data values being generated at the local client devices by adding noise to the actual ordinal data values, the noise being added, according to the first ordering scheme, to each actual data value being probabilistically computed such that a probability of noise being added decreases as an amount of added noise increases;
aggregating, at the remote server device, the first perturbed ordinal data values to at least partially cancel out aggregate noise of the aggregated first perturbed ordinal data values at a population level;
generating, at the remote server device, a second ordering scheme for the ordinal data values by denoising and refining the aggregated first perturbed ordinal data values;
sending, to local client devices, the second ordering scheme;
receiving, at the remote server device from local client devices, second perturbed ordinal data values, the second perturbed ordinal data values resulting from actual non-ordinal data values being accessed, the actual non-ordinal data values being converted to actual ordinal data values according to the mapping, and the first perturbed ordinal data values being generated at the local client devices by adding noise to the actual ordinal data values, the noise being added, according to the second ordering scheme, to each actual data value being probabilistically computed such that a probability of noise being added decreases as an amount of added noise increases;
aggregating, at the remote server device, the second perturbed ordinal data values to at least partially cancel out aggregate noise of the aggregated second perturbed ordinal data values at a population level;
analyzing, at the remote server device, using CLDP, the aggregated second perturbed ordinal data values to identify a computer security threat to the local client devices while preserving privacy of the actual data values by preventing the remote server device from inferring the actual data value of any of the local client devices; and
in response to identifying the computer security threat, protecting against the computer security threat by performing, at one or more of the local client devices, a remedial action to protect the one or more local client devices from the computer security threat.
16. The remote server device of claim 15 , wherein the performing, at the one or more local client devices, of the remedial action comprises one or more of blocking the one or more local client devices from accessing a network, rolling back one or more changes at the one or more local client devices that were made in response to the computer security threat, or temporarily freezing activity at the one or more local client devices, or some combination thereof.
17. The remote server device of claim 15 , wherein the noise is added to each actual ordinal data value with a probability of noise being added decreasing exponentially as an amount of added noise increases.
18. The remote server device of claim 15 , wherein:
the computer security threat comprises a malware outbreak;
the actual non-ordinal data values comprise names of operating systems being executed on the local client devices; and
the adding of the noise to the actual ordinal data values prevents the remote server device from inferring the name of the operating system being executed on any of the local client devices.
19. The remote server device of claim 15 , wherein:
the computer security threat comprises a malware outbreak; and
the analyzing comprises analyzing, at the remote server device, using CLDP, the aggregated second perturbed ordinal data values to identify a timing of a malware infecting the local client devices.
20. The remote server device of claim 15 , wherein the method further comprises:
accessing, at each of the local client devices, a sequence of actual non-ordinal data values;
generating, at each of the local client devices, a sequence of perturbed non-ordinal data values by adding noise to one or more of the actual data values in the sequence of actual non-ordinal data values, an amount of noise being added, according to the mapping and the second ordering scheme, to each actual data value being probabilistically computed such that a probability of noise being added decreases as an amount of added noise increases
sending, from each of the local client devices, the sequence of perturbed non-ordinal data values to the remote server device;
aggregating, at the remote server device, the sequences of perturbed non-ordinal data values to at least partially cancel out aggregate noise of the aggregated sequences of perturbed non-ordinal data values at a population level;
analyzing, at the remote server device, using CLDP, the aggregated sequences of perturbed non-ordinal data values to identify a second computer security threat to the local client devices while preserving privacy of the sequences of actual non-ordinal data values by preventing the remote server device from inferring the sequence of actual non-ordinal data values, and an actual length of the sequence of actual non-ordinal data values, of any of the local client devices; and
in response to identifying the second computer security threat, protecting against the second computer security threat by performing, at one or more of the local client devices, a second remedial action to protect the one or more local client devices from the second computer security threat.