Explanation of Demo for ASCII Code Data

Go to Demo for ASCII Code Data

Explanation of Our Concealing-Restoring System

The following is the explanation of our concealing-restoring system for the physical layer data, and the introduction of its application to the some data on a upper layer of the OSI reference model. Our concealing-restoring system is based on the stochastic filtering theory.

ABSTRACT

We propose a concealing-restoring system (CRS) for data on physical layer of the OSI reference model. CRS conceals those data by disturbing them with some random noises, and restores the data from the concealed ones to the original ones by using the noise elimination based on a proper stochastic filtering theory. Although we introduced the outline of the almost linear version of CRS in our previous work [Fujii & Hirokawa, 2022a], we explain its details, and study its nonlinearization to improve the security of CRS in our paper [Fujii & Hirokawa, 2022b]. We make some theoretical studies of accuracy of CRS in the paper [Hirokawa, 2022] as well as the above one.

INTRODUCTION

The practical use of microdevices for the Internet of Things (IoT) interfaces has made remarkable advance in recent years. Due to its current cutting-edge technologies, IoT including the brain-machine interface (BMI)/ brain-computer interface (BCI) has been turned into the reality. For instance, Benabid et al succeed in controlling an exoskeleton by brain signals of a tetraplegic patient through an epidural wireless BMI. Neuralink reports the news that they have been developing the N1 Link, a fully-implanted, wireless, high-channel count BMI chip. Neuralink White Paper is here. Flesher et al experimentally show that tactile percepts of signals from a robotic, prosthetic arm can be evoked by using a BCI by establishing an afferent channel to the BCI to mimic sensory input from the skin of a hand. It is naturally feared that someone hacks into some embedding type medical devices and hijacks them. The serious apprehension may be beginning to become a reality. It is reported by the US Government Accountability Office (GAO) that a cardiac pacemaker device can be tampered from remote place by radiocommunication. A demonstration of hacking a live jellyfish and the controlling its neural signals is performed by Xu and Dabiri. These are becoming increasingly alarming problems, and we must establish the security in the microdevices. In addition to the security problems above, there are some other problems for the drone (i.e., the flying IoT system in our real life): the hijack of the drone operation, and the exploitation of data on it. We should mind that someone can tap and steal signals between a drone and its remote controller.

We are interested in the security for data in the space which has too small arithmetic capacity to install an encryption technology. The scenes we envision also include countermeasures for the firmware attack and side-channel attack in a low layer of the computer architecture. The firmware attack bypasses some softwares for antivirus and encryption on the higher-layer, and infects the lower-layer data in a device. The side-channel attack bypasses the cryptographic technique based on mathematical complexity and taps the cryptographic key. Several sorts of side-channel attacks have been proposed, and many new side-channel attacks have been presented. In particular, CacheBleed and TLBleed have come under the industrial spotlight.

We propose a concealing-restoring system (CRS) with some secret common keys for the data on the physical layer of the OSI reference model. Here, OSI is the abbreviation of the open systems interconnection, and the OSI reference model consists of 7 layers: the physical layer, the data link layer, the network layer, the transport layer, the session layer, the presentation layer, and the application layer from the lowest layer to the highest one. We restrict our idea to scenes such as the instances described above, and we do not expect general scenes in wireless communication. Thus, the specification and construction of CRS should be for exclusive use among the device users, not be opened to the public. The secret common keys of CRS must be shared by the device users in advance with another method prior to the use of CRS.

Our target scenes requiring such the security include countermeasures for the firmware attack and side-chanel attack in a low layer of the computer architecture. The firmware attack bypasses some softwares for antivirus and encryption on the higher-layer, and infects the lower-layer data in a device. The side-channel attack bypasses the cryptographic technique based on mathematical complexity and taps the cryptographic key. Several sorts of side-channel attacks have been proposed, and many new side-channel attacks have been presented. In particular, CacheBleed and TLBleed have come under the industrial spotlight from the point of view of computer architecture.

We propose a concealing-restoring system (CRS) with some secret common keys for the data on the physical layer of the OSI reference model. Here, OSI is the abbreviation of the open systems interconnection, and the OSI reference model consists of 7 layers: the physical layer, the data link layer, the network layer, the transport layer, the session layer, the presentation layer, and the application layer from the lowest layer to the highest one.

We restrict our idea to scenes such as the instances described above, and we do not expect general scenes in wireless communication. Thus, the specification and construction of CRS should be for exclusive use among the device users, not be opened to the public. The secret common keys of CRS must be shared by the device users %a transmitter and a receiver in advance with another method prior to the use of CRS.

Many endeavors have studied the security on physical layers using individual, physical properties. In the light of noise, Lai, Gamal, and Poor use the random noise to hide the information of the secret key. Tomaru uses the random noise to generate secret keys in the common-key cryptography (symmetric key cryptography). We also use random noises in our CRS, however, that is because we make the random bit flips (i.e., bit errors) directly in bit words on the physical layer. To the best of the authors' knowledge, the concealing-restoring method with using the random noises directly for the data (i.e., message in terms of cryptography) on physical layer is not established yet. We use some mathematical notions as our secret keys for the noise disturbance and noise elimination in CRS. The secret key generation by random noise is among ours. At first glance, the formation of the equations for CRS looks like a generalization of that for the man-in-the-middle (MiM) attacks. However, how to use of the equations in CRS is different from those in the MiM attacks.

We suppose that we maintain the security of the data on physical layer over a proper period of time by installing CRS on the data link layer. Because the data link layer is situated between the physical layer and the network layer, and administers and controls the relation between those two layers. We handle the data on physical layer from the data link layer; CRS should be simple but effective as much as possible. The data concealing is performed by using the random noise disturbance introduced from the data link layer. The data restoring is achieved by the noise elimination in the data link layer. The introduction of the noise disturbance and the noise elimination are based on a proper stochastic filtering theory. We showed its prototype which is the easiest and simplest CRS based on the linear Kalman filtering theory [Fujii & Hirokawa, 2022a]. In this web site, we briefly report how we can change the sort of the noise disturbance and the noise elimination in our CRS to improve its security ability by introducing and using nonlinearity.

EXAMPLES OF RESULTS by CRS

Our concealing-restoring system processes signals on the physical layer from the data link layer of the OSI reference model (Example 1). Based on this process, using some several transformations on upper laysers of the OSI reference model, we can apply our concealing and restoring methods to other objects such as pictorial images and ASCII codes (Examples 3&4). In the examples below, we set N=2 for the parameter N appearing in our paper. Thus, we have 3(=N+1) concealed signals.

Example 1:

A binary word that we want to conceal is given by
100111101100001110011000101110010101001011011100110011010001011000010101101111101110000111111111110.
Using the D/A transformation that we define, we can get the signal

Using our concealing system, we concel this signal and split it into 3 signals in the following.

You can save or send these concealed data.

Assume that a wiretapper becomes aware that the concealed data are for digital data and knows our A/D transformation in some way. Then, he/she gets a binary word from the individual concealed data as follows:
1001111001000011010100001011100101011110110100010000010101010110100110010011101111000001000111100000
for the 1st concealed data,
1001111001000111010110111011110100111001110100110010010101011010000100011011101011001001001110101000
for the 2nd concealed data, and
0000101110000010110001010100101010000001011011000001001100001111000010001000010101100101001110011110
for the 3rd concealed data. We can see the difference between the original signal and the signal obtained from the above binary data of the concealed signal in the following.

On the other hand, using our restoring system, we get the restored data as

The A/D transformation that we define can give the original binary data from the above signal.

Example 2:

Now, we apply our concealing-restoring system to a digital pictorial image. We use binary data of a digital pictorial image in the ORL Database of Faces, an archive of AT&T Laboratories Cambridge. The data have the greyscale value of 256 gradations (8bit/pixel). The original pictorial image and its signal are obtained as in the below:

Here, the upper bound of t is (92x112)x8=10304x8 and t runs over [0,10304x8]. But, in the picture, we show the signal only for t in [0,200].

Our concealing system gives the concealed data as in the following.

Assume that a wiretapper becomes aware that the concealed data are for a digital pictorial image and knows our several transformations to get it from the concealed signal in some way. Then, he/she gets a pictorial image from the individual concealed data as follows:

Using our restoring system, we have the restored signal:

From this resored signal, we can completely restore the original pictorial image. In the following pictures, the left is the original digital pictorial image, and the right is restored one.

Example 3:

Next, we apply our concealing-restoring system to an analog pictorial image in the Olivetti faces database, where the data of pictorial image are transformed to analogue data from the original one in the ORL Database of Faces, an archive of AT&T Laboratories Cambridge. The data have the greyscale value of 256 gradations (8bit/pixel). The original pictorial image and its signal are as in the below:

Here, the upper bound of t is (64x64)x8 = 4096x8 and t runs over [0,4096x8]. But, in the picture, we show the signal only for t in [0,200].

Our concealing system gives the concealed data as in the following.

Assume that a wiretapper becomes aware that the concealed data are for an analog pictorial image and knows our several transformations to get it from the concealed signal in some way. Then, he/she gets a image from the individual concealed data as follows:

Using our restoring system, we have the restored signal:

From this resored signal, we almost completely restore the original pictorial image. In the following pictures, the left is the original analog pictorial image, and the right is restored one.

EXAMPLES OF RESULTS by CRS

We introduce nonlinearity into our concealing-restoring system which processes signals on the physical layer from the data link layer of the OSI reference model (Example 1). In addition to the process, using some several transformations on upper laysers of the OSI reference model, we can handle the security for other objects such as pictorial images and ASCII codes (Examples 5&6). In the examples below, we set N=2 for the parameter N appearing in our paper. Thus, we have 3(=N+1) concealed signals.

Example 4:

We can improve CRS by introducing nonlinearity into it. For the signal of the original binary word in Example 1, we conceal it by the concealing system with the nonlinearity. Then, we have the concealed data:

If we use the same restoring system as in Example 1 to restore the concealed data, we have the signal as in the following.

This says that we fail in recovering the original signal. We can visually realize this failure. Using our own transformations from the signal to the binary word, we get the binary word
111000011011010010101100010101100000001100101110110111101011000101100111110001000011111011100001111,
which is different from the original one. We need a new restoring system for the nonlinear concealing system. Following the method for how to make it, we can get the restored signal by the new restoring system.

This restored signal gives us the original binary word completely.

Example 5:

We use the signal of the digital pictorial image already used in Example 2. We use the concealing system with the nonlinearity, and then, get the concealed data.

Here we show the signals only for t in [0,200].

We indeed have the signal as in the following picture using the restoring system in Example 2, but we fail in recovering the original signal in the following.

Here, we show the signal only for t in [0,200]. We can make this failure visible by making its pictorial image usig our several transformations to get digital pictorial image from a signal. In the following pictures, the left is the original digital pictorial image, and the right is the pictorial image obtained from the above signal.

If we use our new restoring system to cope with the nonlinearity, we can get the restored signal, and recover the digital pictorial image completely:

In the above graph, we show the restored signal only for t in [0,200]. In the pictures, the left pictorial image is original, and the right is the recovered one.

Example 6:

We use the signal of the analog pictorial image already used in Example 3. We use the concealing system with the nonlinearity, and then, get the concealed data.

Here we show the signals only for t in [0,200].

If we use the restoring system in Example 3, we fail in recovering the original signal as in the following graph.

Here, we show the signal only for t in [0,200]. The visualization of this failure is in the following pictures. The left is the original analog pictorial image, and the right is the pictorial image obtained from the above signal.

Using our new restoring system to cope with the nonlinearity, we can get the restored signal, and recover the analog pictorial image completely:

In the above graph, we show the restored signal only for t in [0,200]. In the pictures, the left pictorial image is original, and the right is the recovered one.

Example 7:

We apply CRS to the pictorial image with bigger pixel than the above pictorial images. We use a pictorial image in the Standard Image Data-Base. The number of its total pixels is 512x512=262,144. The pictorial image and the first part of its binary pulse are in the following.