Isolating internet-of-things (IoT) devices using a secure overlay network

1. A method, comprising:

receiving at a compute server of a distributed cloud computing network, over a tunnel established between a customer-premises equipment (CPE) and the compute server, first traffic from an Internet-of-Things (IoT) device that is connected to the CPE, wherein the first traffic is destined to a first destination;

enforcing an egress traffic policy to determine whether the first traffic is permitted to be transmitted to the first destination, wherein the egress traffic policy specifies that traffic is only allowed to be transmitted to a destination associated with a service of the IoT device, wherein the destination associated with the service of the IoT device is external to the distributed cloud computing network;

determining, based on the enforcing of the egress traffic policy, that the first traffic is permitted to be transmitted to the first destination because the first destination corresponds to the destination associated with the service of the IoT device, and responsive to this determination, transmitting the first traffic to the first destination;

receiving, at the compute server, second traffic that is destined to the IoT device;

enforcing an ingress traffic policy to determine whether the second traffic is permitted to be transmitted to the IoT device;

determining, based on the enforcing of the ingress traffic policy, that the second traffic is not permitted to be transmitted to the IoT device, and responsive to this determination, dropping the second traffic;

receiving at the compute server over the tunnel, third traffic from the IoT device that is connected to the CPE, wherein the third traffic is destined to a second destination;

enforcing the egress traffic policy to determine whether the third traffic is permitted to be transmitted to the second destination; and

determining, based on the enforcing of the egress traffic policy, that the third traffic is not permitted to be transmitted to the second destination because the second destination does not correspond to the destination associated with the service of the IoT device, and responsive to this determination, dropping the third traffic.

2. The method of claim 1 , wherein the tunnel is a generic routing encapsulation (GRE) tunnel.

3. The method of claim 2 , wherein a GRE endpoint at the compute server is an anycast IP address that is shared among a plurality of compute servers such that each of the plurality of compute servers can terminate the GRE tunnel.

4. The method of claim 1 , wherein the tunnel is an IPSec tunnel.

5. The method of claim 4 , wherein an endpoint of the IPSec tunnel at the compute server is an anycast IP address that is shared among a plurality of compute servers, wherein stateful information that allows encryption and decryption of the traffic over the IPSec tunnel is stored in a database that is available to each of the plurality of compute servers.

6. The method of claim 1 , further comprising:

wherein the first traffic has a source IP address that is a private IP address; and

prior to transmitting the first traffic to the first destination, translating the private IP address to a public IP address for transmitting the first traffic to the first destination.

7. The method of claim 1 , wherein the egress traffic policy specifies that the first destination is one of one or more predefined destinations to which traffic from the IoT device is allowed to be transmitted.

8. A non-transitory machine-readable storage medium that provides instructions that, if executed by a processor, will cause said processor to perform operations comprising:

receiving at a compute server of a distributed cloud computing network, over a tunnel established between a customer-premises equipment (CPE) and the compute server, first traffic from an Internet-of-Things (IoT) device that is connected to the CPE, wherein the first traffic is destined to a first destination;

enforcing an egress traffic policy to determine whether the first traffic is permitted to be transmitted to the first destination, wherein the egress traffic policy specifies that traffic is only allowed to be transmitted to a destination associated with a service of the IoT device, wherein the destination associated with the service of the IoT device is external to the distributed cloud computing network;

determining, based on the enforcing of the egress traffic policy, that the first traffic is permitted to be transmitted to the first destination because the first destination corresponds to the destination associated with the service of the IoT device, and responsive to this determination, transmitting the first traffic to the first destination;

receiving, at the compute server, second traffic that is destined to the IoT device;

enforcing an ingress traffic policy to determine whether the second traffic is permitted to be transmitted to the IoT device;

determining, based on the enforcing of the ingress traffic policy, that the second traffic is not permitted to be transmitted to the IoT device, and responsive to this determination, dropping the second traffic;

receiving at the compute server over the tunnel, third traffic from the IoT device that is connected to the CPE, wherein the third traffic is destined to a second destination;

enforcing the egress traffic policy to determine whether the third traffic is permitted to be transmitted to the second destination; and

determining, based on the enforcing of the egress traffic policy, that the third traffic is not permitted to be transmitted to the second destination because the second destination does not correspond to the destination associated with the service of the IoT device, and responsive to this determination, dropping the third traffic.

9. The non-transitory machine-readable storage medium of claim 8 , wherein the tunnel is a generic routing encapsulation (GRE) tunnel.

10. The non-transitory machine-readable storage medium of claim 9 , wherein a GRE endpoint at the compute server is an anycast IP address that is shared among a plurality of compute servers such that each of the plurality of compute servers can terminate the GRE tunnel.

11. The non-transitory machine-readable storage medium of claim 8 , wherein the tunnel is an IPSec tunnel.

12. The non-transitory machine-readable storage medium of claim 11 , wherein an endpoint of the IPSec tunnel at the compute server is an anycast IP address that is shared among a plurality of compute servers, wherein stateful information that allows encryption and decryption of traffic over the IPSec tunnel is stored in a database that is available to each of the plurality of compute servers.

13. The non-transitory machine-readable storage medium of claim 8 , wherein the first traffic has a source IP address that is a private IP address, and wherein prior to transmitting the first traffic to the first destination, translating the private IP address to a public IP address for transmitting the first traffic to the first destination.

14. The non-transitory machine-readable storage medium of claim 8 , wherein the egress traffic policy specifies that the first destination is one of one or more predefined destinations to which traffic from the IoT device is allowed to be transmitted.

15. A compute server, comprising:

a processor; and

a non-transitory machine-readable storage medium that provides instructions that, if executed by the processor, will cause the compute server to perform operations comprising,

receiving at the compute server of a distributed cloud computing network, over a tunnel established between a customer-premises equipment (CPE) and the compute server, first traffic from an Internet-of-Things (IoT) device that is connected to the CPE, wherein the first traffic is destined to a first destination,

enforcing an egress traffic policy to determine whether the first traffic is permitted to be transmitted to the first destination, wherein the egress traffic policy specifies that traffic is only allowed to be transmitted to a destination associated with a service of the IoT device, wherein the destination associated with the service of the IoT device is external to the distributed cloud computing network,

determining, based on the enforcing of the egress traffic policy, that the first traffic is permitted to be transmitted to the first destination because the first destination corresponds to the destination associated with the service of the IoT device, and responsive to this determination, transmitting the first traffic to the first destination,

receiving, at the compute server, second traffic that is destined to the IoT device;

enforcing an ingress traffic policy to determine whether the second traffic is permitted to be transmitted to the IoT device,

determining, based on the enforcing of the ingress traffic policy, that the second traffic is not permitted to be transmitted to the IoT device, and responsive to this determination, dropping the second traffic,

receiving at the compute server over the tunnel, third traffic from the IoT device that is connected to the CPE, wherein the third traffic is destined to a second destination,

enforcing the egress traffic policy to determine whether the third traffic is permitted to be transmitted to the second destination, and

determining, based on the enforcing of the egress traffic policy, that the third traffic is not permitted to be transmitted to the second destination because the second destination does not correspond to the destination associated with the service of the IoT device, and responsive to this determination, dropping the third traffic.

16. The compute server of claim 15 , wherein the tunnel is a generic routing encapsulation (GRE) tunnel.

17. The compute server of claim 16 , wherein a GRE endpoint at the compute server is an anycast IP address that is shared among a plurality of compute servers such that each of the plurality of compute servers can terminate the GRE tunnel.

18. The compute server of claim 15 , wherein the tunnel is an IPSec tunnel.

19. The compute server of claim 18 , wherein an endpoint of the IPSec tunnel at the compute server is an anycast IP address that is shared among a plurality of compute servers, wherein stateful information that allows encryption and decryption of the traffic over the IPSec tunnel is stored in a database that is available to each of the plurality of compute servers.

20. The compute server of claim 15 , wherein the first traffic has a source IP address that is a private IP address, and wherein prior to transmitting the first traffic to the first destination, translating the private IP address to a public IP address for transmitting the first traffic to the first destination.

21. The compute server of claim 15 , wherein the egress traffic policy specifies that the first destination is one of one or more predefined destinations to which traffic from the IoT device is allowed to be transmitted.