Harvesting Entropy from Trusted Cryptographic Sources

Information

  • Patent Application
  • 20100023749
  • Publication Number
    20100023749
  • Date Filed
    January 07, 2008
    16 years ago
  • Date Published
    January 28, 2010
    14 years ago
Abstract
Extending entropy in a random number generation utility. Where a device has access to trusted sources of encrypted data, such as encrypted network traffic, such encrypted network traffic may be sampled and the bits fed into the entropy seeding routines of the random number generation utility.
Description
BACKGROUND OF THE INVENTION

The present invention relates to the operation of pseudo-random number generators, and more particularly, to managing and extending entropy pools used in pseudo-random number generators.


Pseudo-random numbers are required for many operations in computers, computer networks, and network devices, and are frequently used in security-critical operations such as cryptographic key derivation. Unpredictability of such random or pseudo-random numbers is paramount for such applications.


Such unpredictability may be defined for example, as a low correlation between bits in a pseudo-random bitstream, between portions of the bitstream, or between the pseudo-random bitstream and other signals.


A good random number utility is a carefully designed and vetted subsystem in a computing device. It may be implemented entirely in software, or it may use specialized hardware assists A typical utility operates on an entropy pool, a pool of randomized bits, from which requests for random numbers may be satisfied.


Personal computers typically derive entropy, which may be defined as a numeric quantity which is unpredictable and uncorrelated from a statistical perspective, from various environmental inputs, particularly those affected by the randomness and unpredictability of human operators Such inputs may include the latency between key presses on a keyboard, idiosyncratic mouse timing, variations in network packet arrival, variation in disk drive operation such as seek time variations, and others. Some systems may devote hardware resources to (pseudo-) random number generation. Values from such sources may be “whitened” by running them through a hashing algorithm such as SHA1 or other cryptographic process, and the resulting bits “stirred” into the entropy pool of the random number utility.


Embedded devices, such as network devices and wireless access nodes, often lack hardware random number generators and are often at a loss for sufficient entropy sources, as such devices do not have unpredictable users, keyboards, mice, disk drives, and the like. Often the “best” response in such devices is to add a hardware generator and use this generator to periodically re-seed the entropy pool used by the random number utility.


What is needed is a way of adding entropy, particularly in embedded systems.





BRIEF DESCRIPTION OF THE DRAWINGS

The invention may be best understood by referring to the following description and accompanying drawings that are used to illustrate embodiments of the invention in which:



FIG. 1 shows a network device operating in a network, and



FIG. 2 shows a flowchart of the invention.





DETAILED DESCRIPTION

Embodiments of the invention relate to extending entropy in random number generation. In an embodiment of the invention, trusted sources of encrypted data, such as encrypted network traffic, are sampled and fed to the entropy seeding routines of the random number generation utility such that the extracted data is stirred into the entropy pool of the random number utility.



FIG. 1. shows an environment suitable for practicing the invention. Network device 100 connects 210 to network 200. Network 200 may be a wired or a wireless network such an Ethernet network operating under known IEEE 803.2 or 802.11 standards. Device 100 operates under control of central processing unit (CPU) 110, which controls one or more network interfaces 120, and operates in conjunction with memory hierarchy 130. In one embodiment, CPU 110 is a MIPS-class CPU such as those from Cavium or Raza. CPUs from other manufacturers, such as Intel, AMD, ARM, or the like may also be used. Network interface 120 may be one or more wired or wireless interfaces. Wired interfaces conforming to IEEE 803.2 standards may be used, or wireless networks conforming to IEEE 803.11 standards. Memory hierarchy 130 as understood by the art holds instructions and data necessary for practicing the invention on machine readable media and typically comprises a small amount of permanent storage for system initialization, fast read-write storage such as DRAM, and bulk storage such as Compact Flash for storing files.


Network 200 is populated by other network devices such as controller 300 and nodes 310 and 320. Nodes 310 and 320 may be other network devices, or they may be peer nodes, such as peers to node 100 in a mesh network.


As understood by the art, such a hardware platform as shown in FIG. 1. operates under the control of an embedded operating system, which may be a LINUX derivative, or other suitable operating system for embedded devices.


If suitable encryption keys are used, the output of robust encryption algorithms is effectively random. According to the present invention, if a device has access to a source of encrypted data, that encrypted data may be sampled and fed into the entropy seeding routines of the random number generation utility, stirring the data into the entropy pool.


A network device 100 such as an access node, switch, or mesh point, has access to multiple trusted sources of encrypted data in the form of encrypted datastreams available to or passing through the device.


When such a device is locally terminating sessions encrypted using WPA or WPA2 protocols, the device is encrypting packets destined for clients, and can sample bits from such packets,


When encryption over the network is centrally terminated, for example at a controller and the encrypted traffic tunneled, a device such as an access node decapsulates the encrypted IEEE 802.11 frames and forwards them over the air; such encrypted 802.11 frames can be sampled.


Mesh nodes involved in forwarding traffic may recognize elements of such traffic as encrypted, and can sample bits from such traffic.


For the purposes of the invention, trusted encrypted packets include those packets which are encrypted by the device itself, or packets which originate from a trusted source, which can include encrypted traffic from peers or controllers which can be trusted not to behave in a malicious manner, and encrypted traffic for which the device holds decryption keys.


Alternatively, encrypted packets may be developed by taking network packets, which may be encrypted packets, and applying further encryption to such packets. In one embodiment, an access node may use a session key from another session to encrypt a packet, producing a packet suitable for use by the invention.


In accordance with an embodiment of the invention and as shown in the flowchart of FIG. 2, a network device recognizes encrypted network traffic and samples bits from such network packets 144, feeding such bits to the entropy seeding routines of the random number utility and its entropy pool 150.


As shown in FIG. 1, packet buffer 140 in memory 130 typically contains packet header data 142, such as TCI/IP headers, and packet data 144. Other encapsulating data, such as Ethernet headers and trailers, are not shown. Packet data 144, which has been identified as encrypted data, is sampled, and added to entropy pool 150.


For the purpose of description herein, it is assumed that the random number utility has an entropy pool, a mechanism for adding or stirring bits into the entropy pool, and a mechanism for returning random bits based on the entropy pool. The nature of these mechanisms will depend necessarily on the computer languages and structures used for their implementation. Such utilities are described, for example, in ANSI X9.31 Appendix A.2.4, or NIST Special Publication (SP) 800-90, incorporated herein by reference.


Such sampling may be event based, as an example, controlled by a timer, or triggered by a low-water mark on the entropy pool. Sampling may take place on a periodic basis,


Sampling may also take place on a stochastic basis. As an example, when encrypted data is recognized, the decision on whether or not to sample that packet is made on the basis of a random number from the random number utility. In one embodiment, a random bit is requested from the random number utility, and if this random bit is a “1” then the packet is sampled, and if “0” the packet is not sampled.


When the decision has been made to sample a packet, either a fixed or a variable portion of encrypted data 144 from the packet may be sampled. In one embodiment, a fixed amount of encrypted packet data starting at a fixed offset from the beginning of packet data 144 is selected and added to the entropy pool. As an example, the offset may be zero, and the entire encrypted packet 144 may be selected and added to the entropy pool. Or, a single byte or fixed number of bytes may be selected and added.


In another embodiment, an offset and length are chosen stochastically, using random numbers from the random number utility. A random number (a number of bits) is requested for an offset into encrypted packet 144 and a random number is requested for the length (number of bits or bytes) to select from encrypted packet 144. The offset should be calculated modulo the packet size, The length should also be calculated modulo the packet size. If the offset plus the length is greater than the length of encrypted packet 144, then the length may be truncated, or data may be sampled from encrypted packet 144 in a circular fashion. The selected data from encrypted packet 144 is then added to the entropy pool.


While the invention has been described in terms of various embodiments, the invention should not be limited to only those embodiments described, but can be practiced with modification and alteration within the spirit and scope of the appended claims. The description is this to be regarded as illustrative rather than limiting.

Claims
  • 1. A method of adding entropy to the entropy pool of a random number generator in a network connected device comprising: identifying a packet of network data as encrypted,selecting the packet for use,extracting data from the packet, andstirring the extracted data into the entropy pool of the random number generator.
  • 2. The method of claim 1 where the network is a wired network.
  • 3. The method of claim 2 where the wired network operates according to IEEE 802.3 standards.
  • 4. The method of claim 1 where the network is a wireless network.
  • 5. The method of claim 4 where the wireless network operates according to IEEE 802.11 standards.
  • 6. The method of claim 1 where the step of selecting the packet for use further comprises selecting a packet on a periodic basis.
  • 7. The method of claim 1 where the step of selecting the packet for use further comprises selecting a packet based on a random number.
  • 8. The method of claim 1 where the step of extracting data from the packet further comprises selecting a fixed amount of data from a fixed position in the packet.
  • 9. The method of claim 1 where the step of extracting data from the packet further comprises: generating a first random number indicating an offset into the packet,generating a second random number indicating a length, andextracting data from the packet starting at the offset generated and for the length generated.
  • 10. The method of claim 9 where the offset and length are used modulo the packet size.
  • 11. The method of claim 1 where the identified packet of network data has been re-encrypted.
  • 12. The method of claim 1 where the identified packet of network data is from a peer node.
  • 13. The method of claim 1 where the identified packet of network data is from a controller.
  • 14. A machine readable medium having a set of instructions stored therein, which when executed on a network connected device cause a set of operations to be performed comprising: identifying a packet of network data as encrypted,selecting the packet for use,extracting data from the packet, andstirring the extracted data into the entropy pool of the random number generator.
  • 15. The machine readable medium of claim 14 where the network is a wired network operating according to IEEE 802.3 standards.
  • 16. The machine readable medium of claim 14 where the network is a wireless network operating according to IEEE 802.11 standards.
  • 17. The machine readable medium of claim 14 where the step of selecting the packet for use further comprises selecting a packet on a periodic basis.
  • 18. The machine readable medium of claim 14 where the step of selecting the packet for use further comprises selecting a packet depending on a random number.
  • 19. The machine readable medium of claim 14 where the step of extracting data form the packet further comprises selecting a fixed amount of data from a fixed position in the packet.
  • 20. The machine readable medium of claim 14 where the step of extracting data from the packet further comprises: generating a first random number indicating an offset into the packet,generating a second random number indicating a length, andextracting data from the packet starting at the offset generated and for the length generated.
  • 21. The machine readable medium of claim 20 where the offset and length are used modulo the packet size.
  • 22. The machine readable medium of claim 14 where the identified packet of network data has been re-encrypted.
  • 23. The machine readable medium of claim 14 where the identified packet of network data is from a peer node.
  • 24. The machine readable medium of claim 14 where the identified packet of network data is from a controller.