JPH11164130A - Method for preventing tampering of image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To effectively prevent digital image data from being changed or tampered with. SOLUTION: (1) A digital image is divided into two areas A and B, the area B is divided into the two areas B1 and B2 from the feature extraction of the data of the area A, then a numerical value H1 obtained from a hash function is calculated from the entire data of the area A, and the area B2 and the numerical value H1 is encoded and embedded in the area B1. (2) Similarly to (1), a numerical value H2 obtained from the hash function is calculated from the entire data of the area A and the area B2, then a numerical value H3 obtained from being decoded from the area B1 is obtained, the numerical values H2 and H3 are compared, and the presence/absence of the change of image data are detected.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for preventing data from being altered or tampered with on a digital screen, and more particularly, to a method for making it difficult to access numeric data embedded in a digital image taken or to prevent the digital data from being altered. The present invention relates to an image change detection method having a function of determining the presence or absence of an image change and capable of effectively preventing digital image tampering.

[0002]

2. Description of the Related Art As digitization of photography has progressed in recent years, photography has been used in a wide range of fields,
It will be compiled as an important report and as an accurate record. Conventionally, the preparation of report documents using photographs requires a lot of man-hours and time since it is manual work from photographing to editing.However, in response to these, personal computers have been Editing of report documents and work records by applying digital image processing technology has become practical. In other words, it is possible to efficiently create a report document using high-quality digital photographs in a short period of time, and without requiring any skill, by using a personal computer to efficiently perform all the editing processes from image capturing to photo pasting. Done.

FIG. 4 is a flowchart for explaining a procedure for creating a report document by digital image processing of a personal computer. First, an object is photographed by a digital camera. The captured image is read at a high speed, digitally converted by a conversion means such as a CCD element, and then converted to M × N.
Digital image data of the pixel is obtained. The obtained digital image data is processed by an image processing system and then temporarily stored in a hard disk system or a storage medium. If necessary, image processing is added, or management information is added and stored.

When editing a digital image, automatic editing for each editing task is selected, and the digital image is automatically pasted into each image-incorporated area on the ledger pasting format every time the image to be pasted is completed. Attached. In addition, when a record document is arbitrarily created, it is possible to simply paste a digital document in accordance with the classification of the record and paste it on the ledger, thereby completing the ledger editing. Then, if the created ledger data is printed out, the report document and the record document are completed.

[0005]

In the above-described image processing method using a personal computer system, a taken image is automatically processed. Therefore, when falsified image data is input, some preventive measures are taken. Otherwise, the final output photo will be the falsified image itself. for that reason,
If the information on the subject is digitized in accordance with the development of digital image processing technology, it becomes an important issue to detect data alteration of the digital image and prevent tampering in the digital image. In addition, since the digital image data can be accessed on a pixel-by-pixel basis, measures corresponding to these are required.

Conventionally, various proposals have been made regarding prevention of image falsification. For example, in Japanese Patent Application Laid-Open No. Hei 6-22119, "an image for preventing illegal outflow of a document by copying by embedding additional information for identifying a user in an image portion of a document and making it readable by a specific extracting means. A processing device "has been proposed. Japanese Patent Application Laid-Open No. 8-30759 discloses that "Fourier transform data of a security management image is encoded in the form of a set of points and embedded in a document image, thereby falsifying information useful for security management of the document. A document security management method which cannot perform the above-mentioned security management image by performing the inverse Fourier transform, which cannot be performed, has been proposed. However, even with the devices and methods proposed above, it cannot be said that this is the best means for detecting data alteration of digital images and preventing tampering.
New improvements need to be made.

According to the present invention, with further development of the digital image processing technology described above, an image change detection which is strongly demanded and which reliably detects data change of a digital image and exhibits an excellent effect of preventing falsification is provided. It is intended to provide a method.

[0008]

As a means for detecting falsified data when a digital image is falsified, there is a method using a one-way hash function.
The one-way hash function is a function for compressing data of an arbitrary length to a fixed length. When different image data is input, the data is converted into different one hundred and several tens of bits of compressed data. However, the hash function has a one-way characteristic because compressed data cannot be restored to data of an arbitrary length. As a typical example of such a hash function,
MD5 (Message Digest 5), SHA (Secure Hash Algri
thm).

When a hash function is used to detect falsified data, a target digital image is input to the hash function, and numerical values obtained from the hash function (hereinafter, simply referred to as “hash values”) are managed. Will be. Specifically, whether or not the image data has been changed or tampered is determined by inputting the target digital image to the same hash function, obtaining a hash value, and comparing the hash value with the managed hash value. be able to.

In the conventional management of a hash value, a method of storing the digital image separately from a target digital image or storing the digital image in a header or the like of the digital image has been used. But,
Regardless of the storage method, the management of the hash value is maintained separately from the data of the digital image itself, making it easy to access the hash value and storing it as the image data is changed or tampered with. The hash value to be changed can be easily changed.

Therefore, in order to prevent alteration or falsification of image data, it is necessary to embed and store the hash value so that access to the hash value becomes difficult. Also, if a change in image data changes the stored hash value immediately or develops a method that destroys the stored state of the hash value, the data change can be easily detected, and the tampering of the digital image is prevented. It will be extremely difficult. Further, even if the image is locally falsified, it is possible to effectively prevent data change and falsification of the digital image if it can be reliably detected.

The present invention has been completed on the basis of the above findings, and has as its gist the following methods (1) and (2) for preventing image falsification.

(1) As shown in FIG. 1, in order to prevent tampering of a digital image, the digital image is divided into two regions A and B, and a region B is divided into two regions B1 by extracting features of the data of the region A. And calculating a numerical value H1 obtained from a hash function from all data in the area A and the area B2, encoding the numerical value H1 and embedding the numerical value H1 in the area B1. It is. Hereinafter, the image tampering prevention method described in (1) is referred to as a first method.

(2) As shown in FIG. 3, upon detecting a change in the data of the digital image, the digital image is divided into two regions A and B, and the region B is divided into two regions based on the feature extraction of the data in the region A. After division into B1 and B2, a numerical value H2 obtained from the hash function is calculated from all the data in the region A and the region B2, and a numerical value H3 obtained by decoding from the region B1 is obtained. H2 and H3
And detecting whether or not the image data has been changed by comparing the image data with the image data. Hereinafter, the image tampering prevention method described in (2) is referred to as a second method.

In the first method, it is difficult to access the embedded hash value, and in the second method, the change of the image data is immediately linked to the change of the embedded hash value. Can be easily detected. Therefore, by combining the above first and second methods, an excellent falsification detection function can be exhibited, and falsification of a digital image can be effectively prevented.

In the present invention, the digital image data is not limited to a specific image format at the time of storage. For example, in addition to the RAW format for storing image data as it is, the JPEG format, T
The IFF format is also included. Further, the image format may be either compressed or uncompressed. That is, an image compressed in the JPEG format is processed by the method of the present invention on the compressed data. Therefore, the following description does not limit the types of image formats.

[0017]

DESCRIPTION OF THE PREFERRED EMBODIMENTS As described above, a one-way hash function compresses data of an arbitrary length into a hash value of one hundred and several tens of bits (128 bits). The same value may occur, so-called "collision". However, when specific image data is changed or falsified to obtain another image data, the hash values of the image data before and after these changes and falsification hardly become the same value stochastically. Therefore, by checking the preceding and following hash values, it is possible to determine whether the image data has been changed or falsified.

Further, independently of the target image data,
Alternatively, the method of holding the hash value has a disadvantage that the access to the hash value is facilitated as described above. Therefore, if the hash value is embedded in the target image itself, the embedding of the hash value itself changes the image data, so that it becomes difficult to access the hash value for falsification of the image data. For example,
If the falsified hash value is embedded, the image data changes. Therefore, the hash value is naturally different from the hash value embedded before the falsification of the image data. Therefore, also in this case, the presence or absence of tampering can be determined based on the hash value.

In the first method of the present invention, an area in which a hash value is to be embedded is specified from target image data.
Then, a hash value is obtained from image data of an area other than the specified area, and stored in a part specified as an embedded area. By employing such a configuration, the hash value can be embedded in the digital image itself as image information that is part of the image data.
For this reason, it is difficult to access the hash value embedded in the image itself.

Next, in the second method of the present invention, a digital image is divided into two regions from target image data, and a hash value is calculated from the specified region. After that, decoding is performed from an area other than the specified area to extract a hash value, and the hash value is compared and confirmed, so that the presence or absence of a change in the image data can be easily detected.

Hereinafter, the specific contents of the first and second methods of the present invention will be described with reference to flowcharts.

FIG. 1 is a flowchart illustrating a method for preventing digital image falsification by the first method of the present invention. First, the digital image is divided into two regions A and B, and the two regions may be continuous or discontinuous.

FIGS. 2A to 2D are diagrams for explaining the configuration of each stage of the target image data. FIG. 3A is a conceptual diagram showing a part of the image data, and shows that the image data is composed of a unit pixel. FIG. 2B is a diagram showing a state in which the region A and the region B are divided into a lattice. A specific example of dividing a digital image into two regions A and B is shown in FIG.
In addition to the method of dividing into a vertical or horizontal grid as shown in
Any processing method can be used as long as the pixels in the area A and the area B can be made to correspond one-to-one, such as a method of dividing the image into three primary colors of a color image simply or vertically or horizontally.

Next, the location where the hash value is to be embedded from the feature extraction of the area A is determined in the area B, and the area is divided into two areas. Here, the feature extraction refers to a process of calculating a feature amount of a measurement target from image data.

FIG. 2 (c) is a diagram showing a state in which the hash value is divided into an area B1 in the area B and an area B2 other than the area B1. Here, as means for dividing the area B into two areas B1 and B2 based on the calculation of the feature amount of the area A, the pixels of the area A and the area B are made to correspond one-to-one, and the M A method of obtaining a variance value at × N pixels and dividing it into two by a threshold, and assigning the same. Similarly, a method of finding an average value at M × N pixels and dividing into two and assigning it to a threshold, or A method of calculating and assigning the pattern similarity is used. In any case, any processing method may be used as long as the feature amount is calculated from the image data of M × N pixels and can be digitized.

FIG. 2D shows M × N pixels in an area A from which a feature is extracted when the embedding position is determined.
Here, the feature amount of the image of the area A may be calculated using the data of the area of M × N pixels excluding the storage location of interest and the mask shown in FIG. Here, the M × N pixels in the region A do not need to be rectangular, and a mask region can be provided therein to form an arbitrary shape. In addition, as an embedding area, an area sufficient to store a hash value by a threshold is secured.

Then, the hash value H1 of the entire image of the region A and the region B2 in which the embedding is not performed in the region B is calculated, and finally, the hash value H1 is embedded in the specified region B1. Here, as a function used when calculating the hash value H1 from all the data in the area A and the area B2, the area B
This is a data string excluding 1, and this arbitrary length data string is MD
5 system, SHA system, AR / DFP system, RIPEMD
A one-way hash function that can compress arbitrary-length data such as (80, 128, and 160 bits) into a fixed length corresponds to this.

When the hash value H1 is encoded and embedded in the area B1, if there is a space to be embedded, the hash value is repeatedly input into the area, or the bit length of the hash value, the size of the embedding area, etc. are encoded. It is desirable to store the data so that the embedded area does not remain.

As a method of encoding the hash value H1 and embedding it in the area B1, n
If the bit value is 0, the value to be written is replaced with an even value close to the original data. If the bit value is 1, the value to be written is replaced with an odd value close to the original data. Just write it to

As described above, in the first method, the access to the hash value is more difficult than in the method in which the hash value is separated from the image data and held, and even if the data is locally falsified. This can be prevented.

FIG. 3 is a flowchart illustrating a method for detecting data change of a digital image according to the second method of the present invention. As is apparent from FIG. 3, the digital image is divided into two regions A and B, and the region B is divided into two regions B1 and B2 from the feature extraction of the data of the region A. The operation up to the calculation of the hash value H2 obtained from all the data of B2 is the same as in the first method described above.

In the second method, the numerical value H3 is extracted by decoding from the area B1. First, the embedding location is specified by the same processing as the method, and the bit is expanded by performing bit expansion. The numerical value H3 can be extracted. And finally, the hash value H2 and the numerical value H3
If they are different, it is determined that the image data has been changed and falsified.

[0033]

As the digitalization of photographs has progressed and the editing of digital screens has become easier, the need for prevention of digital image tampering has increased.
According to the image tampering prevention method of the present invention corresponding to such a request, access to the hash value becomes more difficult than in a method in which the hash value is separated from the image data and stored, and the change of the image data becomes difficult. Since a method of immediately changing the stored hash value or destroying the holding state of the hash value is adopted, data change can be easily detected, and falsification of the digital image becomes extremely difficult. In addition, even if the image is locally falsified, it is possible to reliably detect the falsification, so that an excellent effect of preventing data falsification and falsification of the digital image is exhibited.

[Brief description of the drawings]

FIG. 1 is a flowchart illustrating a method for preventing falsification of a digital image according to a first method of the present invention.

FIGS. 2A and 2B are diagrams illustrating a configuration of image data to be processed at each stage, wherein FIG. 2A is a conceptual diagram illustrating a part of image data, and FIG. FIG. 3C shows a state where the hash value is embedded in the area B, and FIG. 4C shows a state where the hash value is divided into other areas B2. FIG. M × N pixels in the area A where the extraction is performed are shown.

FIG. 3 is a flowchart illustrating a method for detecting data change of a digital image according to a second method of the present invention.

FIG. 4 is a flowchart illustrating a procedure for creating a report document by digital image processing of a personal computer.

Claims (2)

[Claims] To prevent tampering of a digital image, the digital image is divided into two areas A and B,
From the feature extraction of the data of the area A, the area B is divided into two areas B1
And B2, then calculate a numerical value H1 obtained from the hash function from all the data in the area A and the area B2,
A method for preventing image falsification, characterized in that the numerical value H1 is encoded and embedded in the area B1. 2. When detecting a change in the data of a digital image, the digital image is divided into two regions A and B, and a region B is divided into two regions B1 and B2 from the feature extraction of the data of the region A. , Calculating a numerical value H2 obtained from the hash function from all the data in the region A and the region B2, and obtaining a numerical value H3 obtained by decoding from the region B1, and comparing these numerical values H2 and H3. An image tampering prevention method characterized by detecting whether or not image data has been changed.