Patent Application
20070133789
The present invention relates to a compiling device and to a method for compiling as is in particular used for security-relevant fields.
It is possible by hardware errors or most different attack methods to interfere with the function of a processor. Attack methods of this kind may, for example, be realized using light or by probing. Plenty of protection mechanisms which are to prevent an attacker or error to cause an incorrect flow of programs are provided in software or hardware.
It is the object of the present invention to provide a compiling device, a method for compiling and a processor allowing improved program flow protection.
This object is achieved by a compiling device according to claim 1, a compiling method according to claim 14, and a processor according to claim 16.
The present invention provides a compiling device for generating a second program sequence from a first program sequence, the first program sequence comprising a first subarea of a first security level and a second subarea of a second security level distinguishable from the first subarea by at least an indicator, comprising:
recognizing means for recognizing the first subarea of the first program sequence and the second subarea by means of the at least one indicator;
selecting means for selecting instructions from a set of instructions of the second program sequence to map a functionality of the first program sequence to the second program sequence, the set of instructions comprising instructions of a first security category and instructions of a second security category, and the selecting means being formed to select only instructions of the first security category for mapping a functionality of the first subarea and to select instructions of the second security category or the first and second security categories for mapping a functionality of the second subarea; and
generating means for generating the second program sequence from the instructions selected by the selecting means.
In addition, the present invention provides a method for generating a second program sequence from a first program sequence, comprising the steps of:
providing the first program sequence, the first program sequence comprising a first subarea of a first security level and a second subarea of a second security level distinguishable from the first subarea by at least an indicator;
recognizing the first subarea of the first program sequence and the second subarea by means of the at least one indicator;
selecting instructions from a set of instructions of the second program sequence to map a functionality of the first program sequence to the second program sequence, the set of instructions comprising instructions of a first security category and instructions of a second security category, wherein only instructions of the first security category are selected for mapping a functionality of the first subarea and instructions of the second security category or the first and second security categories are selected for mapping a functionality of the second subarea; and
generating the second program sequence from the instructions selected by the selecting means.
The present invention is based on the finding that there are areas in a program which are more critical as to security than other areas. The security classification may take place from a security point of view and from a point of view of considering efficient performance of attacks to a program flow. Exemplarily, the first security level according to the present invention may be associated to an area of the first program sequence which necessitates high security requirements due to a critical functionality, and the second security level may be associated to a subarea which necessitates lower security requirements due to an uncritical functionality.
The second program sequence may, for example, be a sequence of assembler instructions. From a security point of view, the instructions of the second program code, that is, for example, the assembler instructions, are not security-critical to the same extent. Although in principle any instruction may be interfered with anyhow, such as, for example, by spiking or radiation, an attacker will rather tackle deterministic attack scenarios as long as they can be performed efficiently, since this results in easy reproducibility. This is by far a more practical way for attacks seeking a certain result. Typically, deterministic attacks of this kind can be triggered in a well-controlled manner by light, such as, for example, local light. Primarily, regular structures, such as, for example, a register file RAM, are the target of a local attack. If two different jump instructions, such as, for example, a relative jump, like PC=PC+20, and an indirect jump, like PC=REGISTER, are considered, the following situation will arise. With a relative jump where the jump target is determined by the OP code, a successful deterministic attack, like by local light, is considerably less critical than in an indirect jump. This results from the fact that well-aimed manipulating a register by means of light is much easier than altering an OP code of an instruction. As far as the present invention is concerned, a relative jump instruction is exemplarily classified as an uncritical instruction of the first security category and an indirect jump as a critical instruction of the second security category.
The invention is thus based on the finding that it is of advantage to cluster both a program sequence and a set of instructions from a security point of view. Clustering of this kind may, for example, be performed by, on the one hand, classifying the program sequences into critical code and uncritical code and classifying, on the other hand, the instruction set into critical instructions and uncritical instructions:
According to the inventive approach, the usage of critical instructions is limited to those code areas considered to be uncritical. Thus, it is for example prevented in a manner supported by tools that critical instructions are used in critical code areas, increasing general protection against attackers for this area. The approach used so far where there was no security differentiation between the instructions of a microcontroller is thus abandoned.
The inventive method allows a selective classification, controllable by the customer, of parts of the program code and the data. The inventive measure will not represent absolute protection against attacks to program codes or individual instructions. However, by using uncritical instructions, the obstacle for an attack to a critical routine is increased. The inventive limitation of the functionality enhances security against local attacks for a customer.
Preferably, the inventive method can be combined with other measures, such as, for example, EDC, to offer additional protection from global and deterministic attacks. Attacks of this kind may, for example, be provoked by spikes.
Preferred embodiments of the present invention will be detailed subsequently referring to the appended drawings, in which:
a,b are volume representations of sets of instructions according to embodiments of the present invention; and
The first program sequence 112 may be a high-level language program, such as, for example, a C program. The first program sequence 112 contains the description of a certain functionality. The second program sequence 114 may be a machine program executable by a processor. Exemplarily, the second program sequence 114 may represent an assembler program. The compiling device 100 is formed to generate the second program sequence 114 from the first program sequence 112. Thus, the first program sequence 112 is mapped to the second program sequence 114 such that the second program sequence 114 contains the same functionality described in the first program sequence 112.
According to the inventive approach, the first program sequence 112 may be divided into subareas to which different security requirements are associated. Exemplarily, a first subarea of a first security level may be an encryption algorithm or decryption algorithm and a second subarea of a second security level may be a subarea defining a data reception routine, exemplarily from an UART. In this case, executing the encryption/decryption algorithm is more security-critical than executing the data reception routine. Correspondingly, the first subarea has a critical functionality having high security requirements and the second subarea has an uncritical functionality having lower security requirements. The first program sequence 112 may comprise indicators to identify the subareas and/or the security levels associated to the subareas. Exemplarily, a first indicator may be associated to the first security level and a second indicator may be associated to the second security level. The indicator may be a mark, such as, for example, in the form of a word, an instruction or an individual bit, which is exemplarily arranged at the beginning of a subarea and is evaluated by the compiling device 100. Alternatively, each individual instruction of a subarea of the first program sequence may be marked by a corresponding indicator, or else only one respective special subarea, such as, for example, the first subarea, is marked, but not another subarea. The first program sequence 112 may comprise a plurality of subareas which may be first subareas of a first security level, second subareas of a second security level or other subareas of other definable security levels.
According to this embodiment, the first program sequence 112 is provided to both the recognizing means 102 and the selecting means 104. The recognizing means 102 is formed to recognize the security levels of the respective subareas. This may take place by means of the indicators. According to this embodiment, the recognizing means 102 provides a selection signal to the selecting means 104 indicating to the selecting means whether a subarea received in the first program sequence 112 is a first subarea of the first security level or a second subarea of the second security level.
The selecting means 104 is formed to select instructions from a set of instructions. The set of instructions thus is a set of instructions on which the second program sequence 114 is based. If the second program sequence 114 is a machine code which may run on a processor, the set of instructions will be the set of instructions of this processor. The instructions selected map the functionality of the first program sequence 112 to the second program sequence 114.
According to the inventive approach, the instructions of the set of instructions are classified into different security categories. According to this embodiment, the set of instructions comprises instructions of a first security category and instructions of a second security category. The instructions of the first security category may be secure instructions which are, for example, less susceptible to deterministic attacks than the instructions of the second security category. Depending on whether the subarea of the first program sequence 112 is a first subarea of the first security level or a second subarea of the second security level, the selecting means 104 will select either instructions of the first security category or instructions of the second security category. For mapping a functionality of the first subarea of the first security level, the selecting means 104 will, according to the inventive approach, select exclusively instructions of the first security category. For mapping a functionality of the second subarea of the second security level, either instructions of the second security category only or instructions of the first and second security categories may be selected.
The instructions selected for mapping the first program sequence 112 are provided by the selecting means 104 to the generating means 106. The generating means 106 is formed to generate the second program sequence 114 from the instructions selected. Exemplarily, the generating means 106 may be formed to output the selected instructions received in the received order as the second program sequence 114. Alternatively, the selected instructions may be transformed by the generating means 106 to a format required for the second program sequence 114.
The compiling device shown in
In addition, using the security levels allows implicit alteration of the performance of an element in a manner transparent for the customer by means of tool support to enhance the security standard. This allows implicit support of hardware features. If the hardware system on which the second program sequence 114 is executed includes a cache which can only be protected against local attacks insufficiently but supports a write through mechanism, using keywords or indicators may be used to implicitly trigger secure behavior for this module. In the case of a cache, this might be not to store critical data in the cache but to pass on and process the data directly to the CPU. This implicit protection by means of a keyword would, for example, result in special control of the cache by means of SFR (special function register) instructions. This, however, would be transparent for a user who is thus less dependent on the platform. This only entails a loss in performance and/or an increase in the overall energy consumption of the application when this data not arranged in the cache is accessed frequently.
In the programming language C, the code area to be protected may be identified by a special keyword.
Exemplarily, Void _secure ENCRYPT (char *data).
The keyword _secure ensures that, when compiling this function, only assembler instructions considered to be secure are used. It could also be conceivable for the assembler instructions to be classified into several security classes. This corresponds to finer a clustering of both the program sequence and the set of instructions. In this case a security level may for example be required for a certain function.
Exemplarily, Void _secure_level (3) ENCRYPT (char *data).
To allow support of special hardware functions, the compiling device shown in
In a second method step 204, the subareas of the first program sequence are recognized. If a first subarea is recognized, selecting instructions of the first security category will take place in a third step 206 to represent a functionality of the first subarea. If, however, this is a second subarea, selecting instructions of the second security category or the first and second security categories will take place in a parallel third step 208 to realize the functionality of the second subarea.
In a fourth step 210, the second program sequence is generated from the instructions selected in steps 206, 208. The method shown in
a and 3b show embodiments of sets of instructions according to the present invention. According to the embodiment shown in
Alternatively, the set of instructions shown may be divided into further subsets. This allows finer grading of the security categories. In such a case, the first program sequence may comprise further subareas of further security levels identified by further indicators. Additionally, the recognizing means shown in
According to the embodiment shown in
Depending on the circumstances, the inventive method for generating a second program sequence from a first program sequence may be implemented in either hardware or software. The implementation may be on a digital storage medium, in particular on a disc or CD having control signals which may be read out electronically which can cooperate with a programmable computer system such that the corresponding message will be executed. Generally, the invention thus also is in a computer program product having a program code stored on a machine-readable carrier for performing the inventive method when the computer program product runs on a computer. Put differently, the invention may thus also be realized as a computer program having a program code for performing the method when the computer program runs on a computer.
According to an embodiment, the compiling means 100 shown in
| Number | Date | Country | Kind |
|---|---|---|---|
| 102005048029.2 | Oct 2005 | DE | national |
