Method for Generating Private Keys in Wireless Networks

Abstract

The first and second nodes in a wireless network estimate first and second channel response. The first node quantizes the first channel response to produce a first bit sequence, and a feed-forward message, which is transmit as a feed-forward message to the second node. The second node quantizes the second channel response using the feed-forward message to produce and an estimate of the first bit sequence, a second bit sequence and a feed-back message, which is transmitted to the first node. Then, the first and second nodes delete bits in the respective bit sequences using the feed-back and feed-forward message to generate first and second private keys with low bit mismatch rate.

Description

FIELD OF THE INVENTION

This invention relates generally to wireless communications, and more particularly to encrypting and decrypting data using private keys.

BACKGROUND OF THE INVENTION

Security in Wired and Wireless Networks

In wired communication networks, signals are mostly confined in physical transmission media, such as conductive wires and optical fibers. Hence, an eavesdropper can only access the signals by physically attaching to the media.

In contrast as shown in FIG. 1, signals 105 in a wireless network 100 are broadcast by a transmitter node (Alice) 101 over wireless channels. The signals are accessible to any receiver node, e.g., a legitimate user Bob 102 and an eavesdropper Eve 103, within a certain radio propagation range 110. Therefore, security is an extremely important issue in wireless communications.

Public-Key Cryptography

To achieve secure wireless communication, public key cryptography is widely used. The asymmetric structure of the keys does not require a perfectly secure wireless channel to exchange a pair of keys between the transmitter and the receivers. The pair of keys includes a private key and a public key. To operate the network, a public key infrastructure (PKI) generates, distributes and maintains the public keys, in which a trusted party, i.e., a certificate authority (CA), binds all the public keys with a receiver identity and issues a public key certificate to the receiver.

To establish secure wireless communication, the transmitter first verifies the public key certificate of the receiver. After the public key is verified, data are then encrypted using the receiver's public key. The data can only be decrypted using the corresponding private key.

However, for many wireless networks, access to the PKI is difficult, or completely unavailable. In such cases, secure communication in wireless networks becomes challenging. Given this, guaranteeing security in a wireless network is of great interest to users of wireless networks.

Channel Reciprocity

FIG. 2 shows the concept of conventional wireless channel reciprocity 200. The principle of channel reciprocity states that channels between two wireless nodes are reciprocal, i.e., the channel 201 from Alice 101 to Bob 102 has substantially the same characteristics as the channel from Bob to Alice 202, i.e., H_ab=H_ba, for a predetermined frequency range at a predetermined time instance.

The reciprocity of wireless channels enables two nodes to generate private keys based on the channel responses of the reciprocal channels. However, due to noise, interference and hardware impairments, the channels are not perfectly correlated. Therefore, private keys generated independently by the nodes do not always match. If a conventional cryptography technique is used for data to be transmitted, the two keys must be identical, or otherwise the receiver cannot decrypt the data correctly.

Therefore, it is desired that each key has a low bit mismatch rate (BMR). The BMR is defined as a ratio between the number of mismatched bits to the total number of bits in each key.

SUMMARY OF THE INVENTION

Embodiments of the invention provide a method for a first wireless node (user) and a second wireless node (user) to independently generate respective private keys and that have a low bit mismatch rate (BMR) by quantizing correlated signal sources and using public message exchanges.

Each wireless node first independently estimates the channel response H to the other node. The first node quantizes the channel response to produce a first bit sequence and a feed-forward message, and transmits the feed-forward message to the second node via a public channel.

The second node quantizes the channel response based on the received feed-forward message. The second node generates a second bit sequence and a corresponding feed-back message, which is transmitted to the first node via the public channel. Based on the feed-forward and the feed-back message, the nodes deleted selected bits in the first and second bit sequences to produce respective private keys.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic of a conventional wireless network;

FIG. 2 is a schematic of conventional reciprocal wireless channels;

FIG. 3 is a flow chart of a method for generating private keys according to embodiments of the invention;

FIG. 4 is a flow chart of quantization process according to embodiments of the invention; and

FIG. 5 is a flow chart of a universal quantization process according to embodiments of the invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

Method Overview

As shown in FIG. 3, the embodiments of our invention provide a method for generating private keys to encrypt and decrypt data to be transmitted in a wireless network. The method is described by the steps taken by a first node (user Alice) 101 and a second node (user Bob) 102.

During channel sounding and estimating 301, Alice 101 and Bob 102 estimate the response of the channel between the nodes. In one embodiment of the invention, Alice and Bob each transmit a known sounding signal S. When Alice transmits S, Bob receives S and estimates the channel parameters. When Bob transmits S, Alice receives S and estimates the channel parameters.

Due to the noise and hardware impairment, the received signals by Alice and Bob can be expressed respectively as follows:

R _a =H _ba *S+Z _a, and R_b=H_ab*S+Z_b,

where * is a convolution operator, N_ab 201 and H_ba 202 can be vectors, with elements correspond to the channel response for a predetermined transmitter, receiver, frequency, and time. The channel reciprocity states that H_ab=H_ba, and Z_a and Z_b denote noise.

The sounding signals S transmitted by Alice and Bob can also be different as long as the signals are known at the receiver. The sounding signal S can be transmitted multiple times to improve the signal-to-noise ratio (SNR).

Based on the received signal, Alice and Bob estimate the channel independently: The estimated channels are _ba and _ab respectively as

_ba =H _baη_a,

_ab =H _abη_b,

where η_a and η_b represent the channel estimation errors at Alice and Bob, respectively.

Then, Alice and Bob determine channel parameters X 311 and Y 321 as a function of the channel estimates _ba and _ab. The parameters can be simply _ba and _ab, or extracted parameters such as the channel gains, phases at predetermined frequencies, statistical parameters associated with the channel gains, or combinations thereof. If the total number of measurements is n, then the channel parameters X and Y are two length-n bit sequences Xⁿ and Yⁿ, respectively, which can be expressed as:

X ⁿ =f(_ba)=G+F_A,

Y ⁿ =f(_ab)=G+F_B,

where G˜(0,P), F_A˜(0, N_A) and F_B˜(0, N_B) denote the extracted parameter from noise-free channel responses, the estimation error at Alice and the estimation error at Bob, respectively. For simplicity, we consider the case of N_A=N_B=N.

Alice quantizes 302 the bit sequence X using, e.g., a scalar equiprobable quantizer. The level of quantization L is pre-determined. The boundaries for an L=2^l-level quantization are (q₀, q₁], (q₁, q₂], . . . , (q_L−1, q_L). These are selected such that

$Q\left(q_i\right)\stackrel{\Delta }{=}{\int }_{q_i}^{\infty }\frac{1}{\sqrt{2\pi }}\exp \left(-\frac{x^2}{2}\right)x=\frac{L-i}{L},$

The quantizer further partitions each interval (q_i-1, q_i] into M=2^m sub-regions to generate an m-bit feed-forward message based on the bit sequence X. Therefore, the input channel response is quantized into l+m bits, where l bits are in the original bit sequence and m bits are part of the feed-forward message V_a.

Alice transmits 303 the m-bit feed-forward message V_a to Bob.

After Bob receives the feed-forward message V_a, Bob quantizes 305 the bit sequence Y using, e.g., a Maximum a posteriori Probability (MAP) equalizer, which is expressed as

T _bi=argmaxPr(βX_iε(q_j,q_j+1)|Y_i=y_i,V_ai=v_ai),

for i=1, 2, . . . , n, where β is a power normalization factor. The MAP quantization process also produces a feed-back message V_b, which is transmitted 306 to Alice.

In the final steps 307, both Alice and Bob produce respective private keys 313-314 by deleting bits from the original sequences using the feed-back and feed-forward messages.

Feed-Forward and Feed-Back Message

The generation of feed-forward and feed-back bits is illustrated in detail in the following.

FIG. 4 shows details of the quantization processes 302 and 305 by Alice and Bob, respectively. The quantization level L and the feed-forward level M. are pre-determined parameters. In one embodiment of the invention, both Alice and Bob use the scalar equiprobable quantizer designed for a zero-mean and unit-variance Gaussian random variable, i.e., for an L-level quantizer, the quantization boundary q_i to indicate L intervals, (q₀, q₁], (q₁, q₂], . . . , (q_L−1, q_L), is selected such that

$Q\left(q_i\right)\stackrel{\Delta }{=}{\int }_{q_i}^{\infty }\frac{1}{\sqrt{2\pi }}\exp \left(-\frac{x^2}{2}\right)x=\frac{L-i}{L},$

for any iεZ_L, where Q(•) denotes a Gaussian tail function and Z_L={0, 1, . . . , L−1} denotes an integer set. As an example, we set q₀=−∞ and q_L=∞. Gray coding is used for mapping the quantizer indices to bits, e.g., if L=4, the four quantization intervals are mapped to 00, 01, 11, and 10, respectively as shown in FIG. 4.

For generating the feed-forward message, each quantization interval (q_i−1, q_i] is further partitioned into m sub-intervals, (t_i−1,0, t_i−1,1], (t_i−1,1, t_i−1,2], . . . , (t_i−1,m−1, t_i−1,m], where t_i−1,0=q_i−1 and t_i−1,m=q_i, such that each sub-interval (t_i−1,k, t_i−1,k+1] for kεZ_m has an identical probability of

$Q\left(t_{i-1,k}\right)-Q\left(t_{i-1,k+1}\right)=\frac{1}{\mathrm{mL}}.$

Given each channel estimate X_i and Y_i for any iε={1, . . . , n}, Alice and Bob individually quantize the estimates into log₂(L)-bit indices K_A,i and K_B,i, using the L-level scalar equiprobable quantizer. Note that the quantization is performed with a power normalization to have unity variance for quantizing data. The quantized data at iε N are then given by

K _A(i)={j:βx_iε(q_j,q_j+1]}

K _B(i)={j:βy_iε(q_j,q_j+1]}.

After n observations, Alice and Bob obtain bit sequences K_A=[K_A(1), K_A(2), . . . , K_A(n)] and K_B=[K_B(1), K_B(2), . . . , K_B(n)], respectively. Both are n×log₂(L)-bit long.

Alice also generates the log₂(m)-bit feed-forward message V_a(i) from X_i such that V_a(i) is the sub-interval index of the interval K_A(i), more specifically, we can write V_a(i) as follows:

V _a(i)={jβx_iε(t_KA(i),j,t_KA(i),j+1]}.

Alice transmits the n×log₂(m)-bit feed-forward message V_a=[V_a(1), V_a(2), . . . , V_a(n)] to Bob, where each sub-interval index V_a(1) can be expressed by its binary natural code representation with log₂(m) bits.

After receiving the feed-forward message V_a from Alice, and his own observation Yⁿ, Bob carries out a Maximum a Posteriori Probability (MAP) estimation of K′_A. In the MAP estimation, for each Y_i=y_i and V_ai=v_ai (for any iε), Bob searches for the index j_i1 such that the following condition is satisfied:

$j_{i1}=\underset{j\in L_{}}{\mathrm{argmax}}\mathrm{Pr}\left(\beta X_i\in \left(q_j,q_{j+1}\right)Y_i=y_i,V_{\mathrm{ai}}=v_{\mathrm{ai}}\right).$

With the MAP estimate j_i1 and the original quantized data K_B(i), Bob generates the feed-back message V_b(i) which is set to be one if j_i1 and K_B(i) do not match (i.e., j_i1≠K_B(i)), and zero otherwise, for each iε. The feed-back message V_b=[V_b(1), V_b(2), . . . , V_b(n)] of n bits is transmitted to Alice.

Based on the feed-back message V_b, Alice deletes 307 the corresponding K_A(i) if V_b(i)=1, for iε, and sets the remaining as her private key W_A. Similarly, for each iε, Bob also deletes 207 the corresponding bits K_B(i) if V_b(i)=1, and produce his private key W_B using the remaining bits.

The key-generation and deletion of the mismatched bits can be summarized as follows.

• • 1) Alice generates the key K_A based on her received observation.

Alice also generates a feed-forward message V_a based on her received observation and the generated key K_A. Alice sends the feed-forward message V_a to Bob.

• • 2) Bob estimates the key K_B based on his received observation. Bob uses the message V_a and his estimate of the channel response to produce a MAP estimate K′_A of the key K_A.
  • 3) Bob compares his MAP estimate K′_A of K_A with K_B, and generates the feed-back message V_b. The feed-back sequence has value ‘one’ at the locations where the MAP estimate K_A and K_B do not match. At the locations where the MAP estimate K_A and K_B match, V_b has value ‘zero’.
  • 4) Bob sends the feed-back message V_b to Alice. Bob also deletes the bits in his key K_B at locations where V_b has value ‘one.’ The remaining bits of K_B are sets as his private key.
  • 5) Alice deletes the bits in her key K_A at locations where V_b has value ‘one.’ The remaining bits of K_A are sets as her private key.

Universal Quantization

In another embodiment of the invention as shown in FIG. 5, a universal quantization is used, which does not require prior knowledge of the distributions of X and Y.

Alice receives the input channel parameters X 311, and sorts 501 the channel parameters Xⁿ=[X₁, X₂, . . . , X_n] in an ascending order, and partitions the sorted parameters into L intervals with each interval containing an equal number of the parameters X_(i), and then selects quantizing boundaries in an ascending order to generate 502 a first quantization codebook as Q_A={q₀, q₁, . . . , q_L}, i.e., q₀=X₍₁₎, q₁=X_(└n/L┘), . . . , q_L=X_(n).

Bob similarly sorts 501 his input 321 and generates 502 a second quantization codebook Q_B. Q_A and Q_B are not necessarily identical.

Alice and Bob obtain K_A and K_B respectively by universal quantization 503 of the inputs X and Y using the codebooks Q_A and Q_B, respectively. Alice generates the feed-forward message V_A using the same universal equiprobable quantizer principle, i.e., sorting the parameters within a predetermined interval in ascending order and then partitioning the parameters in M sub-intervals to produce the m-bit feed-forward message V_A(i) for X_(i), based on the corresponding sub-interval

Bob generates the feed-back message V_B as follows. For each iε, using the feed-forward message V_ai=v_ai, Bob determines his estimation of Alice's index by j_i1=jεZ_L, if y_iε( q_j, q_j+1), where

${\stackrel{\_}{q}}_j=\{\begin{array}{cc}\mathrm{median}\left[t_{j-1,v_{\mathrm{ai}}+1},t_{j,v_{\mathrm{ai}}}\right),& 1\le j\le L-2,\\ -\infty ,& j=0,\\ \infty ,& j=L-1.\end{array}$

Then, Bob sets V_b(i)=0 if j_i1=K_B(i) and V_b(i)=1 otherwise, for each iε.

The median in the above equation can be replaced by the mean.

EFFECT OF THE INVENTION

The invention provides a method for generating private keys to encrypt and decrypt data to be transmitted in a wireless network. The keys are generated independently based on sounding signals exchanged between two nodes. The private keys have a low bit mismatch rate (BMR) by quantizing correlated signal sources and using public message interchanges.

By transmitting the feed-back and feed-forward messages between two nodes, the BMR can be reduced, e.g., at 20 dB SNR, the BMR is reduced by 40%. The universal quantizer achieves a similar performance as the MAP equalizer when the sources have Gaussian distribution. The universal quantizer has better BMR performance when the source distribution is uniform.

It is to be understood that various other adaptations and modifications may be made within the spirit and scope of the invention. Therefore, it is the object of the appended claims to cover all such variations and modifications as come within the true spirit and scope of the invention.

Claims

1. A method for generating private keys to encrypt and decrypt data to be transmitted in a wireless network including a first node and a second node, comprising: estimating, in the first node, a first channel response to the second node;estimating, in the second node, a second channel response to the first node;quantizing, in the first node, the first channel response to produce a first bit sequence and a feed-forward message;transmitting the feed-forward message to the second node;quantizing, in the second node, the second channel response using the feed-forward message to produce and an estimate of the first bit sequence, a second bit sequence and a feed-back message;transmitting the feed-back message to the first node;deleting, in the first node, bits in the first bit sequence using the feed-back message to generate a first private key; anddeleting, in the second node, bits in the second bit sequence using the feed-forward message to generate a second private key.

2. The method of claim 1, wherein a bit mismatch rate of the private keys is reduced.

3. The method of claim 1, further comprising: transmitting sounding signals to estimate the first channel response and the second channel response.

4. The method of claim 1, wherein the first channel response and the second channel response are based on channel gains, phases at predetermined frequencies, statistical parameters associated with the channel gains, or combinations thereof.

5. The method of claim 1, wherein the first node uses a scalar equiprobable quantizer for the quantizing, and the second node uses a maximum a posteriori probability equalizer for the quantizing.

6. The method of claim 5, wherein the scalar equiprobable quantizer is performed with power normalization.

7. The method of claim 1, wherein the first node and the second node use universal quantization and codebooks for the quantizing.

8. The method of claim 1, further comprising: determining, in the second node, mismatches between the first bit sequence and the second bit sequence to produce the feed-back message.

9. The method of claim 7, further comprising: sorting first channel parameters Xn=[X1, X2, . . . , Xn] of the first channel response in an ascending order, and partitions the sorted first channel parameters into L intervals wherein each interval contains an equal number of the first parameters X(i);selecting quantizing boundaries in an ascending order to generate a first quantization codebook as QA={q0, q1, . . . , qL}, where q0=X(1), q1=X(└n/L┘), . . . , qL=X(n);sorting second channel parameters Yn=[Y1, Y2, . . . , Yn] of the second channel response in an ascending order, and partitions the sorted second channel parameters into L intervals wherein each interval contains an equal number of the parameters X(i);selecting the quantizing boundaries in an ascending order to generate a second quantization codebook as QB={q0, q1, . . . , qL}, where q0=Y(1), q1=Y(└n/L┘), . . . , qL=Y(n).