CN106503721B - Hash algorithm and authentication method based on cmos image sensor PUF - Google Patents
Hash algorithm and authentication method based on cmos image sensor PUF Download PDFInfo
- Publication number
- CN106503721B CN106503721B CN201610956396.6A CN201610956396A CN106503721B CN 106503721 B CN106503721 B CN 106503721B CN 201610956396 A CN201610956396 A CN 201610956396A CN 106503721 B CN106503721 B CN 106503721B
- Authority
- CN
- China
- Prior art keywords
- image
- detected
- puf
- cryptographic hash
- hash
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
- G06V10/446—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering using Haar-like filters, e.g. using integral image techniques
Abstract
The invention discloses a kind of hash algorithm and authentication method based on cmos image sensor PUF, wherein hash algorithm is the robustness that pretreatment mode is utilized and promotes image, then the main feature of image is extracted in two-dimensional dct domain, feature vector F is generated, the unclonable function PUF of physics for recycling cmos image sensor intrinsic establishes exciter response and maps and encrypt generation response, i.e. cryptographic Hash；Wherein authentication method be by calculate and compare image to be detected and original image respectively corresponding to Hash vector distinguish tampered image and similar image.The present invention provides the hash algorithm and authentication method based on cmos image sensor PUF of a kind of trust (RoT) service for image content authentication, it realizes and does not use symmetric key, it can guarantee the safety of certification, the image content authentication especially suitable for bottom hardware from the root.
Description
Technical field
The present invention relates to a kind of hash algorithms, more particularly to a kind of hash algorithm based on cmos image sensor PUF
And authentication method, belong to hardware security and image content authentication field.
Background technique
Past, more and more manpower operational instrument were eliminated during the decade.They are substituted in some crucial applications
The reason is that these are made decisions on one's own, be easy to the to be maliciously tampered operation of picture material of support system is cheated, thus make have it is tight
The erroneous decision of weight consequence.In order to guarantee the content integrity of digital picture, there has been proposed the calculations of many image Hash (Hash)
Method is used for image authentication.
Perceptual image Hash is that a kind of compact of digital picture based on content is summarized.It extracts and is based on from picture material
The main feature of human-eye visual characteristic is then applied to a kind of one hash vector of one-way function generation.Picture material can lead to
The recipient for crossing authorization carries out authentication, and the cryptographic Hash of original image and the Hash of suspect image are compared with effective key
Value.The hash algorithm of one perceptual image is while detecting that the malice carried out to image distorts operation, for useful figure
As content protection operations also have certain tolerance.
Recently, many image hash algorithms are suggested and successfully realize.
A kind of robust perceptual image Hash is initially proposed for careless watermark；The algorithm by by image from
Scattered cosine transform coefficient projects to the zero-mean with key and generates cryptographic Hash in steady mode at random.
Then, a kind of being also suggested based on content Hash algorithm for image local tampering detection, this image Hash
Algorithm is not influenced by filtering but is highly susceptible to geometric distortion interference.The problem of geometric distortion is interfered can be by extracting small
Constant characteristic point solves under perceptual distortion (rotation, distortion etc.).
A kind of method being recently proposed is that most stable of content in image is identified by using SIFT-Harris detector
Feature, then the local feature of extraction is embedded into the descriptor based on Shape context and generates image hash, by distribution
The accidental projection coefficient of message sink coding (DSC) coded quantization image generates cryptographic Hash.
But attacker still has an opportunity before current all hash algorithms execute by directly attacking sensor section
Point carrys out the content of tampered image.In addition, the confidentiality of shared key is determined since above-mentioned hash algorithm all relies on association key
Its security performance is determined.Key is stored in nonvolatile memory (NVM), and these nonvolatile memories are to hold very much
Vulnerable to (such as reverse-engineering, the data remanent magnetism etc.) of Network Intrusion, once key is cracked, attacker is easy for distorting
Picture material, and correct cryptographic Hash is generated for image to be detected.In addition, even if possessing effective key, it also can not be effectively
Prove the legitimacy of picture material acquiring way.In court, not can prove that the creation source of digital picture evidence be can not
Play decisive role.In view of " fruit of blind-your-eyes " principle, invalid information collect means evidence obtained, which will be excluded, to be examined
Except sentencing, therefore, in computer certification, it is able to demonstrate that the source of digital content is very important with equipment.
Summary of the invention
It is a primary object of the present invention to overcome deficiency in the prior art, provide a kind of letter for image content authentication
Appoint the hash algorithm and authentication method based on cmos image sensor PUF of (RoT) service, realizes and do not use symmetric key, energy
The enough safety for guaranteeing certification from the root, the image content authentication especially suitable for bottom hardware.
In order to achieve the above object, the technical scheme adopted by the invention is that:
A kind of hash algorithm based on cmos image sensor PUF, comprising the following steps:
1) original image is acquired by magazine cmos image sensor, original image is pre-processed, obtain picture
Plain size is the pretreatment image of N × N, and N is natural number；
2) in two-dimensional dct domain, feature extraction is carried out to the pretreatment image of N × N, extracts to obtain feature vector F, and simultaneously
Generate a timestamp t；
3) resetting time before the next acquisition image exposure of camera, using feature vector F and timestamp t as
Shift clock cycle NclkWith excitation C, it is input to the unclonable function PUF of physics of magazine cmos image sensor
Generate cryptographic Hash R.
Hash algorithm of the invention is further arranged to: pre-processing in the step 1) to original image, specifically
For,
1-1) size normalization；
Original image collected is normalized into a pixel size as N × N with to down-sampling by bilinearity difference
Normalized image；
1-2) low-pass filtering；
The normalized image of N × N is filtered using gauss low frequency filter；
1-3) histogram equalization；
By the Luminance Distribution normalization of image after filtering processing.
Hash algorithm of the invention is further arranged to: the pretreatment image to N × N in the step 2) carries out feature
It extracts, specifically,
The pretreatment image of N × N 2-1) is divided into nonoverlapping piece, each piece is made of P × P pixel, i.e., shared n
A block generates, n=N2/P2, wherein N, P are natural number；
2-2) B is expressed as by each piecei(i=1,2 ..., n), a pixel is in block BiSpatial position (x, y) at ash
Degree grade is expressed as Bi(x,y)；Two-dimensional dct is applied to block BiIn, then gray level BiThe DCT coefficient of (x, y) are as follows:
Obtain first coefficient Q of the DCT matrix in each piecei(1,1) is exactly DC coefficient, remaining coefficient is then AC system
Number；
Feature vector F 2-3) extracted from pretreatment image is by each piece of BiIn Qi(1,1), Qi(1,2), Qi(2,1)
And Qi(2,2)The sum of absolute value composition, indicate are as follows:
Fi=| Qi(1,1)+Qi(1,2)+Qi(2,1)+Qi(2,2)|
F={ F1,F2,...,Fi,...,Fn-1,Fn}
Wherein, FiFor block BiExtracted feature vector.
Hash algorithm of the invention is further arranged to: making feature vector F and timestamp t respectively in the step 3)
For shift clock cycle NclkWith excitation C, it is input in the unclonable function PUF of physics of magazine cmos image sensor
Generate cryptographic Hash R, specifically,
3-1) to the block B of each imagei, an excitation C is decoded by address decoder to read a pixel-reset
Voltage value PC；
A shadow address C ' 3-2) is generated using excitation C by a linear shift register LFSR；
3-3) excitation C is served as seed and makes its initialization as inputting to be put into linear shift register LFSR, and shadow
Address C ' is exactly linear shift register LFSR displacement NclkOutput after a shift clock cycle；
Wherein, NclkF in feature vector Fi；
3-4) for given excitation C, each image block passes through its feature FiResponse is generated, so a complete figure
Hash output as n bits will be generated；
Cryptographic Hash R is expressed as R=PUF (C, F), and cryptographic Hash R is the feature F and figure for motivating C, picture material by user
Caused by the random PUF determined as the manufacturing variation of sensor itself.
The present invention also provides a kind of authentication methods based on cmos image sensor PUF, comprising the following steps:
A) according to the hash algorithm above-mentioned based on cmos image sensor PUF, the original image that camera is captured leads to
It crosses in the unclonable function PUF of physics of magazine cmos image sensor and generates original image cryptographic Hash R；
B) image to be detected of original image after transmitting is obtained from communication channel, is passed according to above-mentioned based on cmos image
The hash algorithm of sensor PUF, the unclonable function of physics which is passed through into magazine cmos image sensor
Image to be detected cryptographic Hash R ' is generated in PUF；
C) compare original image cryptographic Hash R and image to be detected cryptographic Hash R ', to authenticate whether image to be detected is original
The true content of image；
If the bit of original image cryptographic Hash R and image to be detected cryptographic Hash R ' within the set range be it is equal,
Success is then authenticated, the true content that the image to be detected is original image is authenticated；
Otherwise, authentification failure, it is believed that the image to be detected is the forgery of original image or distorts version.
Authentication method of the invention is further arranged to: comparison original image cryptographic Hash R in the step c) and to be checked
Altimetric image cryptographic Hash R ' is weighed by calculating the Hamming distance d between original image cryptographic Hash R and image to be detected cryptographic Hash R '
Amount,
Wherein, HD () indicates normalized Hamming distance, i.e., the ratio between the string of binary characters of two equal lengths
Special difference section；For mould 2 plus operation, RiFor block B in original imageiCorresponding cryptographic Hash, R 'iFor block B in image to be detectedi
Corresponding cryptographic Hash；
If d > η, η are to distort threshold value, then it is assumed that the content of the image to be detected has been tampered；Otherwise, then it is assumed that this is to be checked
No change has taken place or only have passed through normal image processing operations for the content of altimetric image.
Authentication method of the invention is further arranged to: thinking that the image to be detected is original image in the step c)
Forgery or distort version, specifically include following several situations,
(1) image to be detected distorts version for original image, and the region being tampered then passes through unmatched bit
The position of position determines；
(2) image to be detected is the forgery version of original image, content be it is true, but the value of timestamp by
Change；
(3) image to be detected is not the camera institute captured image by capture original image；
(4) the case where any combination of three cases above.
Compared with prior art, the invention has the advantages that:
1, dimension normalization, low is mainly utilized in the hash algorithm provided by the invention based on cmos image sensor PUF
The pretreatment modes such as pass filter, histogram equalization promote the robustness of image, then extract the main of image in two-dimensional dct domain
Feature generates feature vector F, and the unclonable function PUF of the physics for recycling cmos image sensor intrinsic establishes excitation and rings
It should map and encrypt generation response, i.e. cryptographic Hash.The cryptographic Hash is calculated in sensor layer, by the physical structure institute of PUF
It determines, the root that a kind of low cost can be provided for digital content source trusts (RoT) service.And the existing image based on key
It is helpless when hash algorithm is the problems such as facing illegal invasion, data remanent magnetism and Replay Attack, but hash algorithm of the present invention
Since Hash vector can only be generated when imaging sensor is run, so for illegal invasion, data remanent magnetism and Replay Attack etc.
Safety problem has immune function.
2, the authentication method provided by the invention based on cmos image sensor PUF passes through calculating and mapping more to be checked
As with original image respectively corresponding to Hash vector distinguish tampered image and similar image, be able to detect and normally handle
Image still passes through the image of content tampering, and the success or not of certification depends on the response and original image that image to be detected generates
Response comparison result.
3, the image with identical Hash vector is generated since two cmos image sensors can not be manufactured, so figure
As cryptographic Hash can be used to identify video generation device；Cmos image sensor PUF is mentioned as a random one-way function to generate
Take feature and timestamp to Hash vector mapping can also in digital image content as evidence of crime in use, at the beginning of generation
The equipment and time of crime of beginning image provide conclusive errorless proof.
Above content is only the general introduction of technical solution of the present invention, in order to better understand technological means of the invention, under
In conjunction with attached drawing, the invention will be further described in face.
Detailed description of the invention
Fig. 1 is that the present invention is based on the flow charts of the hash algorithm of cmos image sensor PUF；
Fig. 2 is that the present invention is based on the cryptographic Hash of the hash algorithm of cmos image sensor PUF to generate block diagram；
Fig. 3 is that the present invention is based on the flow charts of the authentication method of cmos image sensor PUF；
Fig. 4 is original image and image to be detected for being tampered of content in present invention experiment；
Fig. 5 is the normalization Chinese of the image perception cryptographic Hash of 54 image to be detected and original image in present invention experiment
Prescribed distance；
Fig. 6 is to be filtered the image after the average filter operation having a size of 7 × 7 in present invention experiment to original image.
Specific embodiment
With reference to the accompanying drawings of the specification, the present invention is further illustrated.
The present invention provides a kind of hash algorithm based on cmos image sensor PUF, as shown in Figure 1, comprising the following steps:
1) original image is acquired by magazine cmos image sensor, original image is pre-processed, obtain picture
Plain size is the pretreatment image of N × N, and N is natural number.
Wherein, original image is pre-processed, specifically,
1-1) size normalization；
Original image collected is normalized into a pixel size as N × N with to down-sampling by bilinearity difference
Normalized image；Normalized purpose is: not only to aeroelastic requirements of the guarantee hash algorithm to repetitive operation, and in order to
The image hash for ensuring to generate has a fixed length and identical computation complexity.So pixel N is larger, can make to count
Calculation complexity is higher, and N is smaller, can ignore some detail sections on picture.
1-2) low-pass filtering；
The normalized image of N × N is filtered using gauss low frequency filter, can be reduced in communication channel
The influence of noise pollution or filtering.
It is usually obtained by the convolution of the core of dimensional gaussian distribution and image, Gaussian kernel coefficient G (i, j) is from one two
Dimension Gaussian function carries out sampling calculating:
In formula (1), σ is the standard deviation of Gaussian Profile, and i and j are since core origin respectively in trunnion axis and vertical
The integer distance of axis；
The average value of the Gaussian kernel distribution of core is assumed zero, and normalized discrete approximation Gaussian kernel can indicate are as follows:
The size of core should increase σ to keep the Gaussian properties of filter, and one 5 × 5 core just can be used for low in an experiment
Pass filter.
1-3) histogram equalization；
By the Luminance Distribution normalization of image after filtering processing, hash algorithm can be promoted to the robustness of Luminance Distribution.
If I (i, j) represents grey scale pixel value of the input picture at position (i, j)；The image I's of one N × N pixel
Intensity profile is divided into L grades, that is, { X0,X1,...,XL-1,
It is assumed that nkFor gray value X in input picturekSum of all pixels, then XkProbability density indicate are as follows:
In formula (3), p (Xk) and XkRelationship be referred to as probability density function PDF, in fact, histogram is exactly the figure of PDF
Shape appearance.
And the probability density function PDF based on image can calculate cumulative distribution function CDF:
Obvious c (XL-1)=1.
The transfer function of histogram equalization can indicate are as follows:
f(Xk)=X0+(XL-1-X0)cdf(Xk), k=0,1 ..., L-1 (5)
Image I '=I ' (i, j) after then equalizing can be indicated are as follows:
I '=f (I)
={ f (I (i, j)) I (i, j) ∈ Xk,j∈0,1,...,
N-1and k∈0,1,...,L-1 (6)
2) in two-dimensional dct domain, feature extraction is carried out to the pretreatment image of N × N, extracts to obtain feature vector F, and simultaneously
Generate a timestamp t.Low frequency component depicts the baseline of image, and high fdrequency component represents the detail section of sophisticated image；
Therefore, the key message of an image concentrates in several DC coefficients and low frequency coefficient in DCT matrix.
Wherein, feature extraction is carried out to the pretreatment image of N × N, specifically,
The pretreatment image of N × N 2-1) is divided into nonoverlapping piece, each piece is made of P × P pixel, i.e., shared n
A block generates, n=N2/P2, wherein N, P are natural number；
2-2) B is expressed as by each piecei(i=1,2 ..., n), a pixel is in block BiSpatial position (x, y) at ash
Degree grade is expressed as Bi(x,y)；Two-dimensional dct is applied to block BiIn, then gray level BiThe DCT coefficient of (x, y) are as follows:
Obtain first coefficient Q of the DCT matrix in each piecei(1,1) is exactly DC coefficient, remaining coefficient is then AC system
Number；
Feature vector F 2-3) extracted from pretreatment image is by each piece of BiIn Qi(1,1), Qi(1,2), Qi(2,1)
And Qi(2,2)The sum of absolute value composition, indicate are as follows:
Wherein, FiFor block BiExtracted feature vector.
3) resetting time before the next acquisition image exposure of camera, using feature vector F and timestamp t as
Shift clock cycle NclkWith excitation C, it is input to the unclonable function PUF of physics of magazine cmos image sensor
Generate cryptographic Hash R.
Key of the invention is no longer the certainty hard coded that can be saved, shift, but by inquiring for generating image
What the imaging sensor of content directly generated.This is by the built-in intrinsic unclonable function of physics in cmos image sensors
What PUF was completed.Collection have the PUF of IC manufacturing difference can use one group of external input generate one group it is unique, substantially not
The output may be imitated, predicted.
It can be regarded as a black box exciter response system:
R=PUF (C) (9)
Input and output bit stream R and C are called excitation and response respectively.
Due to the defect and uncertainty of manufacturing process, exciter response mapping can not be replicated.Fixed pattern noise
It (FPN) is the output pixel voltage caused by being misplaced due to the change in process of sensor with interconnection under uniform illumination or dark
The variation of value.The excitation of PUF is exactly the address of a pixel pair.Response is exactly its selected pixel pair corresponding to each excitation
Resetting voltage comparison result.For there is sub-fraction output response can be because the variation of environmental condition generates asking for overturning
Topic, can be very good to be solved using correction techniques.
Hash is generated using cmos image sensor PUF, as shown in Fig. 2, specifically,
3-1) to the block B of each imagei, an excitation C is decoded by address decoder to read a pixel-reset
Voltage value PC；
A shadow address C ' 3-2) is generated using excitation C by a linear shift register LFSR；
3-3) excitation C is served as seed and makes its initialization as inputting to be put into linear shift register LFSR, and shadow
Address C ' is exactly linear shift register LFSR displacement NclkOutput after a shift clock cycle；
Wherein, NclkF in feature vector Fi；
3-4) for given excitation C, each image block passes through its feature FiResponse is generated, so a complete figure
Hash output as n bits will be generated；
Cryptographic Hash R is expressed as R=PUF (C, F), and cryptographic Hash R is the feature F and figure for motivating C, picture material by user
Caused by the random PUF determined as the manufacturing variation of sensor itself.
By taking the image block of 8 × 8 10 bits as an example, by two-dimensional dct generate real coefficient section [8 × 8 × (- 512),
8 × 8 × 512] between=[- 32768,32768].Based on the required precision of IEEE IDCT, the precision of 14 bits is just able to satisfy
The needs of cosine coefficient.It therefore, can be by FiQuantify to 16 bits.The length of LFSR should be equal to or more than 16 bits to keep away
Exempt from Hash collision.In all possible situation, excitation C can be an arbitrary number either encrypted random number with
Eliminate the influence of Replay Attack.
The present invention also provides a kind of authentication methods based on cmos image sensor PUF, as shown in figure 3, including following step
It is rapid:
A) according to Fig. 1 and the hash algorithm shown in Fig. 2 based on cmos image sensor PUF, the original that camera is captured
Original image cryptographic Hash R is generated in the unclonable function PUF of physics that beginning image passes through magazine cmos image sensor；
B) image to be detected after transmitting of original image is obtained from communication channel, according to Fig. 1 and shown in Fig. 2 be based on
The image to be detected is passed through the physics of magazine cmos image sensor not by the hash algorithm of cmos image sensor PUF
Image to be detected cryptographic Hash R ' can be generated in cloning function PUF；
C) compare original image cryptographic Hash R and image to be detected cryptographic Hash R ', to authenticate whether image to be detected is original
The true content of image；
If the bit of original image cryptographic Hash R and image to be detected cryptographic Hash R ' within the set range be it is equal,
Success is then authenticated, the true content that the image to be detected is original image is authenticated；
Otherwise, authentification failure, it is believed that the image to be detected is the forgery of original image or distorts version, is specifically included following
The case where several possibilities,
(1) image to be detected distorts version for original image, and the region being tampered then passes through unmatched bit
The position of position determines；
(2) image to be detected is the forgery version of original image, content be it is true, but the value of timestamp by
Change；
(3) image to be detected is not the camera institute captured image by capture original image；
(4) the case where any combination of three cases above.
Wherein, compare original image cryptographic Hash R and image to be detected cryptographic Hash R ', it can also be by calculating original image Hash
Hamming distance d between value R and image to be detected cryptographic Hash R ' measures the similitude between two images,
Wherein, HD () indicates normalized Hamming distance, i.e., the ratio between the string of binary characters of two equal lengths
Special difference section；For mould 2 plus operation, RiFor block B in original imageiCorresponding cryptographic Hash, R 'iFor block B in image to be detectedi
Corresponding cryptographic Hash；
If d > η, η are to distort threshold value (η is an arbitrarily small score), then it is assumed that the content of the image to be detected by
It distorts；Otherwise, then it is assumed that no change has taken place or only have passed through normal image procossing behaviour for the content of the image to be detected
Make.
Since normal image processing operations and malice tampered image content all can change cryptographic Hash in different degree,
And due to perceptual discrimination power and perceptual sensitivity be often it is unknown, inconsistent, under the influence of vision scope, area
Point content keeps operation that can only be determined by experiment with the threshold value η maliciously distorted.
It is relatively described in detail below by way of specific experimental image.
Firstly, acquiring CRP pairs from previously fabricated cmos image sensor；Experiment high speed imaging sensor used
Resolution ratio receives 64 × 64 3t-aps arrays of small size, a column grade CDS, and an on piece 10 ranks grade ADC and one
The limitation of sense buffer.
Then total to have 64 by the image segmentation of 64 × 64 pixels sensed at the block of 8 × 8 pixel non-overlaps
Block.
Calculated feature vector F and timestamp t is fed back to imaging sensor PUF again and generates 64 Hash, output is breathed out
Before uncommon value R, using original image of the MATALAB script to capture carry out pretreatment and in two-dimensional dct domain extraction feature to
Measure F.
Image to be detected that original image and content are tampered, as shown in Figure 4.Wherein, (a) in Fig. 4 is original graph
Picture, (b)~(f) in Fig. 4 are image to be detected that content is tampered.The figure of (b)~(f) in Fig. 4 totally five content tamperings
As by Photoshop software editing.
In addition, 49 counterparts (content is protected) are led on the basis of one group of content protection operations for being listed in table 1
Cross what modification original image obtained.So image to be detected shares 54, including the protected image of 49 contents and 5
The image that content is tampered.
Table 1
And the normalization Hamming distance of 54 image to be detected and the image perception cryptographic Hash of original image, as shown in Figure 5.
As it can be seen that the normalization HDs for image to be detected image that content is tampered is at least 32.8%, and the similar image of most contents
Normalization HDs be lower than 10%.By protection image, device HDs in the case where distortion effect most serious is 29.7% to some contents,
This problem is due to caused by excessive image enhancement.Fig. 6 shows the result that average filter operation is carried out to original image.
The operation seriously reduces the perceived quality of image, and normalization Hamming distance is caused to be up to 28.1%.This is in this experiment very
The example of the noise channel excess contamination occurred less.
In addition, in an experiment, should be selected appropriate to distort threshold according to the image processing operations determined by application requirement
Value η.In this experiment, we can choose η=30% to distinguish content tampering image and content similar image, while we
The value of η be can reduce to enhance the sensibility to distorted image, although this can weaken the robust for benign image processing operations
Property.
Basic principles and main features and advantage of the invention have been shown and described above.The technical staff of the industry should
Understand, the present invention is not limited to the above embodiments, and the above embodiments and description only describe originals of the invention
Reason, without departing from the spirit and scope of the present invention, various changes and improvements may be made to the invention, these changes and improvements
It all fall within the protetion scope of the claimed invention.The claimed scope of the invention is by appended claims and its equivalent circle
It is fixed.
Claims (5)
1. a kind of hash method based on cmos image sensor PUF, which comprises the following steps:
1) original image is acquired by magazine cmos image sensor, original image is pre-processed, it is big to obtain pixel
The small pretreatment image for N × N, N are natural number；
2) in two-dimensional dct domain, feature extraction is carried out to the pretreatment image of N × N, extracts to obtain feature vector F, and generate simultaneously
One timestamp t；
3) resetting time before the next acquisition image exposure of camera, using feature vector F and timestamp t as displacement
Clock cycle NclkWith excitation C, it is input in the unclonable function PUF of physics of magazine cmos image sensor and generates
Cryptographic Hash R；
The pretreatment image to N × N in the step 2) carries out feature extraction, specifically,
The pretreatment image of N × N 2-1) is divided into nonoverlapping piece, each piece is made of P × P pixel, that is, shares n block
It generates, n=N2/P2, wherein N, P are natural number；
2-2) B is expressed as by each piecei, wherein i=1,2 ..., n, a pixel is in block BiSpatial position (x, y) at ash
Degree grade is expressed as Bi(x,y)；Two-dimensional dct is applied to block BiIn, then gray level BiThe DCT coefficient of (x, y) are as follows:
Obtain first coefficient Q of the DCT matrix in each piecei(1,1) is exactly DC coefficient, remaining coefficient is then AC coefficient；
Feature vector F 2-3) extracted from pretreatment image is by each piece of BiIn Qi(1,1), Qi(1,2), Qi(2,1) and Qi
The sum of the absolute value of (2,2) composition, indicates are as follows:
Fi=| Qi(1,1)+Qi(1,2)+Qi(2,1)+Qi(2,2)|
F={ F1,F2,...,Fi,...,Fn-1,Fn}
Wherein,FiFor block BiExtracted feature vector；
In the step 3) using feature vector F and timestamp t as shift clock cycle NclkWith excitation C, it is input to phase
Cryptographic Hash R is generated in the unclonable function PUF of the physics of cmos image sensor in machine, specifically,
3-1) to the block B of each imagei, an excitation C is decoded by address decoder to read a pixel reset voltage value
PC；
A shadow address C ' 3-2) is generated using excitation C by a linear shift register LFSR；
3-3) excitation C is served as seed and makes its initialization as inputting to be put into linear shift register LFSR, and shadow address
C ' is exactly linear shift register LFSR displacement NclkOutput after a shift clock cycle；
Wherein, NclkF in feature vector Fi；
3-4) for given excitation C, each image block passes through its feature FiResponse is generated, so a complete image will
The Hash output of n bits can be generated；
Cryptographic Hash R is expressed as R=PUF (C, F), and cryptographic Hash R is passed by the excitation C of user, the feature F of picture material and image
Caused by the random PUF that the manufacturing variation of sensor itself is determined.
2. the hash method according to claim 1 based on cmos image sensor PUF, it is characterised in that: the step
1) original image is pre-processed in, specifically,
1-1) size normalization；
Original image collected is normalized into a pixel size returning for N × N with to down-sampling by bilinearity difference
One changes image；
1-2) low-pass filtering；
The normalized image of N × N is filtered using gauss low frequency filter；
1-3) histogram equalization；
By the Luminance Distribution normalization of image after filtering processing.
3. a kind of authentication method based on cmos image sensor PUF, which comprises the following steps:
A) hash method according to claim 1 based on cmos image sensor PUF, the original graph that camera is captured
As by generating original image cryptographic Hash R in the unclonable function PUF of physics of magazine cmos image sensor；
B) image to be detected of original image after transmitting is obtained from communication channel, it is according to claim 1 to be based on CMOS
The hash method of imaging sensor PUF, by the image to be detected by the physics of magazine cmos image sensor can not gram
Image to be detected cryptographic Hash R ' is generated in grand function PUF；
C) compare original image cryptographic Hash R and image to be detected cryptographic Hash R ', to authenticate whether image to be detected is original image
True content；
If the bit of original image cryptographic Hash R and image to be detected cryptographic Hash R ' within the set range be it is equal, recognize
Demonstrate,prove the true content for successfully authenticating that the image to be detected is original image；
Otherwise, authentification failure, it is believed that the image to be detected is the forgery of original image or distorts version.
4. the authentication method according to claim 3 based on cmos image sensor PUF, it is characterised in that: the step
C) comparison original image cryptographic Hash R and image to be detected cryptographic Hash R ' in, by calculating original image cryptographic Hash R and to be detected
Hamming distance d between image cryptographic Hash R ' is measured,
Wherein, HD () indicates normalized Hamming distance, i.e., the bit difference between the string of binary characters of two equal lengths
Different part；For mould 2 plus operation, RiFor block B in original imageiCorresponding cryptographic Hash, Ri' it is block B in image to be detectediIt is corresponding
Cryptographic Hash；
If d > η, η are to distort threshold value, then it is assumed that the content of the image to be detected has been tampered；Otherwise, then it is assumed that the mapping to be checked
No change has taken place or only have passed through normal image processing operations for the content of picture.
5. the authentication method according to claim 3 based on cmos image sensor PUF, it is characterised in that: the step
C) in think the image to be detected be original image forgery or distort version, specifically include following several situations,
(1) image to be detected distorts version for original image, and the region being tampered then passes through unmatched bit
Position determines；
(2) image to be detected is the forgery version of original image, and content is true, but the value of timestamp is changed
Become；
(3) image to be detected is not the camera institute captured image by capture original image；
(4) the case where any combination of three cases above.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610956396.6A CN106503721B (en) | 2016-10-27 | 2016-10-27 | Hash algorithm and authentication method based on cmos image sensor PUF |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610956396.6A CN106503721B (en) | 2016-10-27 | 2016-10-27 | Hash algorithm and authentication method based on cmos image sensor PUF |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106503721A CN106503721A (en) | 2017-03-15 |
| CN106503721B true CN106503721B (en) | 2019-07-16 |
Family
ID=58321536
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610956396.6A Active CN106503721B (en) | 2016-10-27 | 2016-10-27 | Hash algorithm and authentication method based on cmos image sensor PUF |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106503721B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106972934A (en) * | 2017-04-28 | 2017-07-21 | 山东大学 | A kind of camera images storage system based on PUF |
| CN107257285B (en) * | 2017-07-20 | 2023-03-10 | 中国工程物理研究院电子工程研究所 | Authentication system based on single photon excitation and optical PUF |
| CN107395369B (en) * | 2017-08-19 | 2021-03-02 | 大家传承网络科技（深圳）有限公司 | Authentication method, access method and system for self-contained equipment of mobile Internet |
| CN107766750B (en) * | 2017-11-22 | 2023-05-09 | 河海大学常州校区 | PUF circuit based on threshold voltage reference |
| CN110858803B (en) * | 2018-08-23 | 2022-10-04 | 中国电信股份有限公司 | Authentication method, system, server, and computer-readable storage medium |
| CN108898191A (en) * | 2018-09-26 | 2018-11-27 | 苏州米特希赛尔人工智能有限公司 | Convolutional neural networks feature-extraction images sensor |
| CN109618124A (en) * | 2018-09-26 | 2019-04-12 | 苏州米特希赛尔人工智能有限公司 | Feature-extraction images sensor |
| CN112712090A (en) * | 2019-10-24 | 2021-04-27 | 北京易真学思教育科技有限公司 | Image processing method, device, equipment and storage medium |
| CN111161259B (en) * | 2019-12-31 | 2021-06-22 | 支付宝(杭州)信息技术有限公司 | Method and device for detecting whether image is tampered or not and electronic equipment |
| CN111460526B (en) * | 2020-04-17 | 2021-10-12 | 支付宝(杭州)信息技术有限公司 | Image data recording, acquiring and verifying method and device based on block chain |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102325131A (en) * | 2011-07-20 | 2012-01-18 | 北京邮电大学 | Bidirectional identity authentication method for wireless sensor network node |
| CN102546149A (en) * | 2012-01-16 | 2012-07-04 | 华南理工大学 | Crypto chip system and secret key extraction method |
| CN103778374A (en) * | 2014-02-19 | 2014-05-07 | 邹候文 | Trusted terminal, double-channel card, anti-cloning chip, chip fingerprint and channel attack resistance method |
| CN103778590A (en) * | 2012-10-24 | 2014-05-07 | 航天信息股份有限公司 | Method and device for utilizing digital image to store and transmit information |
| CN104392207A (en) * | 2014-11-10 | 2015-03-04 | 天津大学 | Characteristic encoding method for recognizing digital image content |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9996480B2 (en) * | 2012-07-18 | 2018-06-12 | Analog Devices, Inc. | Resilient device authentication system with metadata binding |
-
2016
- 2016-10-27 CN CN201610956396.6A patent/CN106503721B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102325131A (en) * | 2011-07-20 | 2012-01-18 | 北京邮电大学 | Bidirectional identity authentication method for wireless sensor network node |
| CN102546149A (en) * | 2012-01-16 | 2012-07-04 | 华南理工大学 | Crypto chip system and secret key extraction method |
| CN103778590A (en) * | 2012-10-24 | 2014-05-07 | 航天信息股份有限公司 | Method and device for utilizing digital image to store and transmit information |
| CN103778374A (en) * | 2014-02-19 | 2014-05-07 | 邹候文 | Trusted terminal, double-channel card, anti-cloning chip, chip fingerprint and channel attack resistance method |
| CN104392207A (en) * | 2014-11-10 | 2015-03-04 | 天津大学 | Characteristic encoding method for recognizing digital image content |
Non-Patent Citations (2)
| Title |
|---|
| CMOS Image Sensor Based Physical UNCLONABLE Function for Coherent Sensor-Level Authentication;Yuan Cao et al.;《IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS_I: REGULAR PAPERS》;20151130;第62卷(第11期);第2.1节第一行，第2.2节，第2.4节 |
| 基于图像正则化和DCT的变换的感知哈希技术;安坤 等;《工业控制计算机》;20140331;第27卷(第3期);第三节A部分，B部分，第4节F部分、图3-4，图13 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106503721A (en) | 2017-03-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106503721B (en) | Hash algorithm and authentication method based on cmos image sensor PUF | |
| Tang et al. | Robust image hashing with ring partition and invariant vector distance | |
| Wang et al. | Detection of image region duplication forgery using model with circle block | |
| Qin et al. | Perceptual image hashing with selective sampling for salient structure features | |
| CN102096894B (en) | Image fragile watermarking algorithm capable of realizing accurate positioning of tampered region | |
| Tomioka et al. | Robust digital camera identification based on pairwise magnitude relations of clustered sensor pattern noise | |
| Pal et al. | Watermarking scheme using local binary pattern for image authentication and tamper detection through dual image | |
| CN104636764B (en) | A kind of image latent writing analysis method and its device | |
| Hou et al. | Detection of hue modification using photo response nonuniformity | |
| Sowmya et al. | Video authentication using spatio temporal relationship for tampering detection | |
| CN104217387A (en) | Image watermark embedding and extracting method and device based on quantization embedding | |
| CN111104872A (en) | GF-2 image integrity authentication method applying SIFT and SVD perceptual hashing | |
| Pal et al. | Robust watermarking scheme for tamper detection and authentication exploiting CA | |
| Cao et al. | Using image sensor PUF as root of trust for birthmarking of perceptual image hash | |
| Ouyang et al. | A semi-fragile watermarking tamper localization method based on QDFT and multi-view fusion | |
| Saikia et al. | Image authentication under geometric attacks via concentric square partition based image hashing | |
| Mushtaq et al. | Novel method for image splicing detection | |
| CN102881008B (en) | Based on the anti-rotation image Hash method of annulus statistical nature | |
| Chaturvedi et al. | A novel approach for data hiding using LSB on edges of a Gray scale cover images | |
| Zhu et al. | Recaptured image detection through enhanced residual-based correlation coefficients | |
| CN102945542A (en) | Region-of-interest authentication and tampering detection digital watermark method | |
| Manjunatha et al. | A survey on image forgery detection techniques | |
| Neelima et al. | Perceptual hash function for images based on hierarchical ordinal pattern | |
| Xiao et al. | A semi-fragile watermarking tolerant of Laplacian sharpening | |
| Doegar et al. | Image forgery detection based on fusion of lightweight deep learning models |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210707 Address after: 213200 No.26 Heshan Ring Road, Jintan Economic Development Zone, Changzhou City, Jiangsu Province Patentee after: JIANGSU WOERSON ELECTRONIC TECHNOLOGY Co.,Ltd. Address before: 213022, No. 200 Jinling North Road, Jiangsu, Changzhou Patentee before: CHANGZHOU CAMPUS OF HOHAI University |