The present invention relates generally to computer communications, and, more particularly, to encryption-based methods for transferring micropayments.
Electronic commerce continues to grow at a tremendous pace. New communications technologies, such as WiFi and WiMax, decrease the costs of providing network services (such as cellular voice services and wireless data services), leading to a greatly increased number of service providers. Previous network models, where a few large central carriers controlled their networks and charged for access to them, are being supplanted by a model including many disparate providers. Commercial and financial models also change. For example, as roaming between service providers becomes more frequent, selection of a carrier might be negotiable on the spur of the moment and may even be negotiable during a call or data session. With a large and ever changing number of service providers, it becomes difficult, if not impossible, for each service provider to establish business relationships with all other service providers. Without these pre-established arrangements to reconcile charges, brokers step forward to handle billing. Service providers work with brokers to get reimbursed for the services they provide, end customers reimburse the brokers, and brokers extract fees for processing the payments.
The increased number of service providers also changes the financial model with respect to end customers. While interacting with these various service providers, a customer makes numerous small payments for service. Traditional methods for reconciling payments (e.g., credit-card systems) are not appropriate to these “micropayments,” because the cost overhead of reconciling each payment would swamp the value of the micropayment itself.
Micropayment systems have been proposed to handle these small, incremental payments in a manner cost-effective both to the end customers and to the vendors. Some of these systems use a cryptographic construct called a “hash chain.” A hash chain is generated by repeated applications of a cryptographic hash function. Each entry in a hash chain is then used to verify a micropayment. A broker verifies the micropayments, reimburses the vendor, and charges the end customer. Because cryptographic hash chains allow a service provider or vendor to aggregate individual micropayments, he saves on transaction costs with the broker. A hash-chain-based system also provides for non-repudiation and prevents fraudulent accounting by service providers and vendors.
The above considerations, and others, are addressed by the present invention, which can be understood by referring to the specification, drawings, and claims. According to aspects of the present invention, micropayments are represented by individual hash-chain members. The hash chains are then aggregated to provide a more efficient data exchange between a vendor and a broker.
In one embodiment, an end user (here called the “payer”) cryptographically signs “commitments” and transmits then to a vendor (i.e., a network-service provider). Each commitment includes an anchor of a hash chain and an “accumulated count” field which tracks the total number of micropayments made thus far in the payment transaction between the payer and the vendor. The payer can also transmit payment tokens to the vendor. Each payment token includes an element of the hash chain, the hash chain being secured by the anchor included in the commitment.
When the vendor seeks reimbursement from a broker, the vendor tells the broker the total number of micropayments in the payment transaction. (The number may be based, for example, on the accumulated count in the last commitment of the payment transaction plus any micropayments made in payment tokens after the last commitment). The vendor need not send every intervening commitment to the broker. This saves on transmission costs between the vendor and the broker and on storage costs for both of them.
In some embodiments, a verification system is established between the broker and the payer. The commitments transmitted by the payer to the vendor include information tied to this verification system. (For example, the verification information can include a timestamp or a counter.) The vendor checks the authenticity of the payer's commitments and micropayments. In turn, the vendor sends verification information to the broker. The broker checks this information against the verification system established with the payer. If the information is verified to be correct, then the broker reimburses the vendor for the services provided and charges the payer. The verification information ensures that the payer and vendor cannot cheat each other by, for example, repudiating legitimate payments or by submitting the same information for multiple reimbursements.
While the appended claims set forth the features of the present invention with particularity, the invention, together with its objects and advantages, may be best understood from the following detailed description taken in conjunction with the accompanying drawings of which:
Turning to the drawings, wherein like reference numerals refer to like elements, the invention is illustrated as being implemented in a suitable environment. The following description is based on embodiments of the invention and should not be taken as limiting the invention with regard to alternative embodiments that are not explicitly described herein.
Many known prior-art systems use a cryptographic hash function to make micropayments. This hash function, h: {0,1}*→{0,1}n, maps a variable-length input to a fixed-length output. It is intended to be a practical realization of a random function. While it is easy to compute, it is very difficult to invert. SHA-1 is a well known example of such a hash function; it produces a 160-bit output. A hash chain of e entries is of the form e0, e1, . . . , ec, ec+1, where e0 is called the anchor of the hash chain, ec+1 is a (virtually) random number, and ei=h(ei+1) for the hash function h( ).
Hash chains were first proposed in the context of one-time passwords and have since been proposed for micropayments. In the context of micropayments, each entry in the hash chain is used as a payment worth some pre-determined amount. Specifically, prior-art micropayment techniques often include the following steps. (These steps, modified as appropriate, are also used in the discussion below to describe embodiments of the present invention.)
Specifically, the broker 104 checks that the signatures, the fields in the commitments, and the hash chains are valid, and that no previously used hash chain has been reused. The broker 104 then reimburses the vendor 102 and bills the payer 100.
By using hash chains for micropayments, several payments fall within the scope of a single signature operation. The vendor 102 benefits because he only has to perform (a) one hash operation for every payment and (b) one signature verification for the initial commitment. Also, the hash-chain payments are aggregated at the vendor 102. That is, if the vendor 102 has already been paid e1, . . . , ei−1, and is then paid ei, then the vendor 102 only needs to store ei and not all of the previous payments. This is because the hash function is assumed to have the property that ei−1 can be generated efficiently only by someone that possesses ei, and furthermore, it is difficult to find some other ej such that h(ej)=ei−1. This aggregation decreases the amount of data that the vendor 102 needs to send to the broker 104 when requesting reimbursement. Similarly, the broker 104 has reduced computation costs as he only needs to verify the signature on every commitment and not on every payment. Hash chains also reduce the computation needed in the device of the payer 100 as not every payment needs to be signed.
However, the prior-art system of
The choice of hash-chain length also affects the vendor 102 and the broker 104. In the example from
In contrast to the prior-art techniques discussed above and illustrated by
Step 1 of an embodiment of the present invention is similar to Step 1 as described above: The payer 100 receives a certificate from a trusted authority which could be, but need not be, the broker 104. (See Step 400 of
Step 2 in the present embodiment can differ from the above described Step 2 in numerous ways. First, a commitment includes three new fields. One field is called the “accumulated count,” a second field is the “verifier,” and a third field is a “transaction identifier.” (Various embodiments exclude one or more of these fields, as discussed in detail below. The present discussion is meant to be broadly illustrative rather than limiting.) Second, Step 2 can be repeated within one payment transaction, that is, a single payment transaction can include multiple commitments.
To illustrate these points, in the prior-art technique of
As in the prior-art technique, for each hash chain, the payer 100 can send payment tokens to the vendor 102, each token including a member of the current hash chain to indicate payment (Step 408 of
In some embodiments, the first commitment in a payment transaction either does not include an accumulated count (in which case it is assumed to be zero), or it includes a non-zero (possibly random) number. These cases are described below in the discussion of Steps 4 and 5.
In some embodiments, the accumulated count allows the commitments to replace some or all of the payment tokens. Because the accumulated count tracks the number of micropayments made in the payment transaction between the payer 100 and the vendor 102, the payer 100 can indicate payments simply by sending the commitments rather than by sending payment tokens. The accounting for payments is discussed below in reference to Steps 4 and 5.
The verifier field is used differently in different embodiments of the present invention. In one embodiment, the verifier is a timestamp that records the relative or actual time when a commitment is made. (In this case, the date field D discussed above may be redundant.) The timestamp is of sufficient granularity that no two commitments in the same payment transaction between the payer 100 and the vendor 102 can have the same value. Furthermore, for two commitments M1 and M2 in the same payment transaction, where M1 is sent before M2, the timestamp in M1 is smaller than the timestamp in M2. Some embodiments use the current time (in GMT, say) to a sufficient granularity for the verifier timestamp. In other embodiments, the verifier field is an ordered counter. The counter is checked to make sure that it always progresses monotonically in a pre-agreed manner (e.g., always increases or always decreases) from one commitment to the next within a given payment transaction.
Some embodiments include a transaction identifier field in each commitment. This is useful if the vendor 102 intends to support concurrent payment transactions with the payer 100. In the prior-art technique, the anchor of the hash chain can serve as a transaction identifier. In some embodiments of the present invention, the anchor of the first hash chain in a payment transaction can work as well, as long as the payer 100 does not attempt to reuse that hash chain.
Calculations predict that 32 bits are sufficient for each of the accumulated count and transaction identifier fields, and 64 bits are sufficient for the verifier. (The 64-bit representation of time in version 4 of the Network Time Protocol, for example, provides a resolution of up to a fraction of a nanosecond.) Consequently, embodiments of the present invention increase the size of each commitment by only 16 bytes (for embodiments that include all three new fields).
Moving on to Step 3, in embodiments of the present invention, the vendor 102 can receive multiple commitments in one payment transaction (Step 500 of
In Step 4, the vendor 102 seeks reimbursement from the broker 104 for the payment transaction. In the prior-art technique of
In Step 5, the broker 104 receives the reimbursement request 306 (Step 600 of
If the reimbursement request 306 is verified to the satisfaction of the broker 104, then the broker 104 calculates the number of micropayments represented by the request 306. (Variations in this process are described above in reference to Step 4.) The broker 104 then translates this number of micropayments into a reimbursement amount (possibly minus a transaction fee) (Step 604 of
The present inventions provides advantages in performance (storage space and processing time) over prior-art techniques. To illustrate these advantages, the following discussion compares an embodiment of the prior-art technique with an embodiment of the present invention. As different embodiments exhibit different performance characteristics, this discussion is illustrative only and is not meant to limit the invention in any way.
For personal communications devices such as cell phones and PDAs, tests indicate that generating a 163-bit ECC curve 3 signature takes roughly 100 times as long as generating a SHA-1 hash of 20 bytes. Also, verifying a signature takes about three times as long as generating the signature. (ECC is preferred over RSA signatures because of the limited computational ability of these personal devices.)
To calculate the time needed for the vendor 102 and the broker 104 to process payments, let p be the number of payments the payer 100 makes, h be the length of a hash chain, and r be the number of reimbursements that have already been processed for the payer 100 by the broker 104. Use the time needed to generate one hash as the unit of time. Let ts be the time needed to generate a signature and tv the time to verify a signature. (As discussed above, ts=100 and tv=300 for a 163-bit ECC curve 3 cryptosystem). In the prior-art technique, the time to process payments from a payer 100 at the vendor 102 and at the broker 104 is then:
T
old
=P+┌p/h┐'(tv+1)
where ┌ ┐ is the ceiling function. The p component represents the number of hashes to be verified. ┌p/h┐×tv represents the number of commitments made by the payer 100 to make p payments and the signatures on those commitments that need to be verified. Finally, ┌p/h┐×1 represents the need to compute the hash of each commitment to compare with the hashes of prior commitments for payment transactions that have already been reimbursed. In contrast, in an embodiment of the present invention, the time to process p payments from the payer 100 at the vendor 102 is:
T
v,new
=p+┌p/h┐×t
v.
The vendor 102 verifies p hashes and ┌p/h┐ commitments (signatures). He also verifies ┌p/h┐ verifiers, but that time is considered to be negligible when compared to the time required for the cryptography-related verifications. As the above formulas for Told and Tv,new suggest, the difference between the processing times at the vendor 102 is attributable to the prior art's need to check against previous commitments. The advantage of embodiments of the present invention grows linearly with the ratio p/h.
In an embodiment of the present invention, the time to process these p payments at the broker 104 (when the vendor 102 files for reimbursement) is:
T
b,new
=c×t
v+(p mod h)+(1+└p/h┘−┌p/h┐)×h
where c is 1 if p≦h, and 2 otherwise, and mod is the modulo operator (the remainder after dividing p by h). The c×tv component comes from the fact that the broker 104 verifies only one commitment if p≦h and two commitments otherwise. The remainder of the expression is the number of hashes that the broker 104 verifies for entries from the hash chain associated with the final commitment. These calculations show that for the broker 104 the difference between the prior-art and present techniques is quite pronounced.
Turning to the payer 100, for a given hash chain length h, the processing time at the payer 100 is the same for the prior-art and the present techniques:
┌p/h┌(ts+h).
The payer 100 makes a tradeoff in choosing the length h of the hash chain. Because the payer 100 is not always able to predict exactly how many payments he will make, he runs the risk of generating a long hash chain and wasting either time or space or both. Embodiments of the present invention provide flexibility because the payer 100 can still choose relatively short hash chains and not waste processing time or space. To quantify the risk from the prior-art technique, consider two hash chain lengths, hs and h1, with h1>>hs. Consider the case where the payer 100 is willing to trade off time for space. If the payer 100 is willing to store at most hs hash-chain entries at one time, then, in the prior-art technique, the payer 100 regenerates hash-chain entries each time hs entries are exhausted. The total processing time at the payer 100 under the prior-art technique in this case is:
In contrast, the processing time for the payer 100 when using an embodiment of the present invention is:
T
u,new
=┌p/h
s┐(ts+hs)
If the payer 100 is willing to trade off space for time, then his space requirements go up commensurately. For example, when h1=10×hs the payer 100 allocates ten times as much space. When hs=10, this is the difference between allocating 200 bytes and 2 megabytes (SHA-1 hashes are 20 bytes each). The latter can be a significant amount of storage to allocate to a single payment session.
The above discussion shows that embodiments of the present invention provide processing-time benefits to the vendor 102 and to the broker 104. For a given hash chain length, the prior-art and present techniques are identical in terms of processing time for the payer 100. However, embodiments of the present invention are still advantageous for the payer 100 because they perform well even with smaller hash chains. Smaller hash chains are beneficial to the payer 100 because he does not risk wasting processing time or storage space.
To compare the prior-art and present techniques from the standpoint of storage requirements at the broker 104, the vendor 102, and the payer 100, let sh be the space needed to store an entry from a hash chain, and let sc be the space needed to store a commitment. When SHA-1 is the hash function, sh is 20 bytes for the hash plus 4 bytes for the index in the hash chain. The size of sc includes the signature, which is about 60 bytes for a 163-bit curve 3 ECC cryptosystem; however, sc includes whatever else is in the commitment, such as the (hash of the) service agreement between the payer 100 and the vendor 102. It is expected that sc is about five times the size of sh.
The space required at the payer 100 for payments is the same in the prior-art and present techniques. The payer 100 needs to store the unspent entries from the hash chain. The waste of space at the payer 100 has a linear relationship to the number of payments he makes. In addition the payer 100 can store receipts for payments he has already made. Under an embodiment of the present invention, a receipt includes <M1, Mn, ei, i>, while under the prior art, a receipt includes <Mj, ei, i>. The space required at the payer 100 for such receipts is quite different for the prior-art vs. the present techniques: For the prior-art, it is: ┌p/h┐×sc+sh, and for a present embodiment it is k×sc+sh, where k=1 if p≦h, and k=2 otherwise. Thus, the storage requirement increases linearly under the prior art but is constant under embodiments of the present invention.
At the vendor 102 and the broker 104, the space required by an embodiment of the present invention is very different from the requirements under the prior art. In a present embodiment, the broker 104 stores only one timestamp once he has reimbursed the vendor 102 for any reimbursement requests. The vendor 102 also stores only a single timestamp for all reimbursements that have been made to him. (Every future payment he accepts from a payer 100 should have a timestamp that is later than this stored timestamp.) Consequently, in the present embodiment, the space required by the vendor 102 for an un-reimbursed payment transaction is k×sc+sh, where k=1 if p≦h, and k=2 otherwise. Under the prior art, the corresponding space requirement is sc×┌p/h┐+sh×(1+r), where r is the number of payment transactions for which the vendor 102 has already been reimbursed. Here, r reflects the fact that the vendor 102 stores (the hash of) previous commitments so that he can check against them to detect any attempts by the payer 100 to double spend. Under the prior-art, these r hashes are also stored at the broker 104 to ensure that the vendor 102 does not attempt to get reimbursed more than once for the same payment transaction.
The prior-art and present techniques are identical for the vendor 102 when p≦2 h and r=0. However, for other values of p the space remains constant under the present embodiment but increases linearly under the prior art. This is shown graphically in
To summarize some of the benefits of embodiments of the present invention over the prior art: The vendor 102 reaps tremendous space and data-transfer benefits. The broker 104 processes less data and stores dramatically less data. The payer 100 uses less storage space for receipts for payments already made.
In view of the many possible embodiments to which the principles of this invention may be applied, it should be recognized that the embodiments described herein with respect to the drawing figures are meant to be illustrative only and should not be taken as limiting the scope of the invention. For example, different known hash and cryptographic signature methods may be used. Therefore, the invention as described herein contemplates all such embodiments as may come within the scope of the following claims and equivalents thereof.