CN109684829A - Service call monitoring method and system in a kind of virtualized environment - Google Patents

Abstract

The invention discloses a method and system for monitoring service invocation in a virtualized environment. The method is: 1) the cloud generates a hash library of key code segments in the monitored guest virtual machine according to the authorized access service program and related source files set by the tenant; 2) after the monitored guest virtual machine is started, the service calls monitoring The monitor sets the permission bits of the memory page loaded by the monitored guest virtual machine, so that the code on the memory page can check the integrity of the key code segment in the memory before execution; 3) The service call monitor detects the service call When it occurs, the page directory address of the calling process is obtained according to the characteristics of the service call. When the page directory address of the calling process is in the whitelist process list, the service call is allowed to be executed; when the return result event of the service call is detected. , obtain the information of the calling process according to the characteristics of the returned result event, and allow the returned result to pass when the calling process is in the whitelisted process list.

Description

A service call monitoring method and system in a virtualized environment

technical field

The invention belongs to the technical field of computer security, and in particular relates to a method and a system for providing service call monitoring in a virtualized environment.

Background technique

Cloud computing applications based on virtualization technology are developing rapidly. With the popularization of cloud computing applications, all services are migrating from local to cloud. According to the RightScale report, 95% of the IT practitioners surveyed said their companies are using cloud computing services. At the same time, in order to attract more tenants, cloud service providers provide many additional services so that tenants can focus on their core business. However, the security of calling additional services provided by cloud service providers currently relies on the user's ID and password. Once the user's ID and password are leaked, the additional services ordered by tenants may be maliciously invoked by adversaries to achieve their malicious purposes. For example, cloud service providers are now deploying key management services and encryption services in the cloud. The key management service isolates the user's key from the guest virtual machine. Even if the guest virtual machine is compromised by an adversary, the adversary cannot obtain the user's key. However, the security of calling cryptographic computing services currently depends on the user's ID and password. Once the adversary obtains the user's ID and password, the cryptographic computing service can be maliciously invoked in the guest virtual machine to achieve the adversary's malicious purpose. Usually, in order to ensure the automatic service of the service, it is also necessary to write the user's ID and password in the configuration file of the customer virtual machine, which greatly increases the security threat. Services are also no longer secure.

The virtual machine monitor is an important component of the virtualization platform, which is responsible for allocating and managing the resources of the host machine, so that the guest virtual machines running on it can share the physical resources on the host machine. Generally, the hardware device used by the guest virtual machine is simulated by the virtual machine monitor, and the virtual machine monitor handles all the exit events of the guest virtual machine.

Virtual Machine Introspection (VMI) is a technology that monitors the running status of a virtual machine outside the virtual machine. The monitoring function of this technology is done by observing memory details, trapping in hardware events and reading CPU registers. Most VMI implementations need to know the operating system knowledge of the guest virtual machine in advance, and use the operating system knowledge and the memory information of the guest virtual machine to analyze the state of the guest virtual machine. However, when the current VMI tool performs introspection on the guest virtual machine, it needs to suspend the guest virtual machine, which introduces a large performance overhead, and the VMI tool relies on the knowledge of the guest virtual machine operating system to parse the semantically readable information. Once the operating system of the guest virtual machine is damaged, the parsed information will no longer be trusted. In the current detection scheme based on virtual machine introspection technology, in order to reduce the performance loss caused by virtual machine introspection, an auxiliary module of the guest virtual machine is required to trigger security detection. However, once the guest virtual machine is compromised by the adversary, the auxiliary modules installed in the guest virtual machine are no longer trusted.

Most of the CPU chips introduced by AMD and Intel support the nested page table technology, which greatly improves the speed of converting the virtual address of the guest virtual machine into the physical address of the host machine through the nested page table technology. While improving performance, the nested page table has permission bit flags. Generally, there are three permission bits: readable, writable and executable. Once a permission violation occurs, an exception is triggered, causing the guest virtual machine to exit, and the virtual machine monitor handles the virtual machine exit event. Among them, if the permission of a memory page is set to agnostic, when the code on the memory page is to be executed, an instruction fetch error will be triggered. Figure 5 shows a schematic diagram of address translation and permission bit setting using nested page tables.

SUMMARY OF THE INVENTION

Aiming at the above-mentioned security problem that service invocation security only depends on user ID and password, the present invention proposes a method and system for providing service invocation monitoring in a virtualized environment. This solution designs a service invocation monitor to detect service invocation events in the guest virtual machine in real time, and to perform real-time detection on the integrity of key codes in the guest virtual machine. Only those with good integrity and authorized are allowed. The program invokes the service and provides the audit function at the same time. The service call monitoring system is implemented as a component of a virtual machine monitor, and the architecture of the service call monitoring system is shown in Figure 1.

Specifically, the technical scheme adopted in the present invention is as follows:

A method for providing service invocation monitoring, the solution provides services in a virtualized environment, and the specific steps include:

1) Generate the operating system fingerprint information of the monitored guest virtual machine for the monitoring system to solve some semantic gap problems; use the process structure and system symbol table in the fingerprint information to construct the process chain information, and traverse all the processes in the process chain to obtain the call The page directory address information of the process;

2) The tenant sets up the programs authorized to access the service in a secure environment, and transmits the program source files authorized to access the service to the cloud through secure network transmission, and the cloud uses these binary source files to generate the monitored client virtual machine. The hash library of key code segments is used by the service call monitoring system to verify the integrity of key code segments;

3) When the guest virtual machine starts, the service call monitor uses the nested page table technology provided in the CPU chip to set the permission bits of the memory page loaded by the guest virtual machine, so that the code on the memory page can be executed in real time before execution. Verify the integrity of key code segments in memory, and maintain a list of whitelisted processes. The whitelisted process linked list stores the page directory address of the whitelisted process; among them, the whitelisted process is authorized and the code segment is complete Sexually verified processes;

4) The service invocation monitor detects the service invocation in the virtual machine monitor VMM, and when it detects the occurrence of a service invocation event, invokes the virtual machine state detection module, and obtains the information of the invoking process according to the characteristics of the service invocation, such as calling through the network. KMS service, the data packets sent to KMS can be found by filtering network data packets to determine that a cryptographic operation invocation event has occurred in the VM. The IP and port of the network connection are both characteristics of the service invocation. The memory performs introspection and obtains the page directory address of the calling process. Only when the page directory address of the calling process is in the whitelist process list, the service call is allowed to pass (that is, the call is allowed to be sent to the server);

5) For a service call with a return result, the service call monitor detects the return result event of the service call in the VMM, and when it detects that the return result event of the service call occurs, it calls the virtual machine state detection module, and according to the return result event Only when the calling process is in the whitelist process list, the result returned by this service call is allowed to pass (that is, the result returned by the server is allowed to be returned to the monitored virtual machine).

Further, the service call monitoring system can be implemented in a Xen-based virtualization system, or in a virtualization system based on VMware ESX/ESXi and Hyper-V, or in a KVM-QEMU-based virtualization system. .

Further in step 1), the user generates operating system fingerprint information of the monitored guest virtual machine in a safe environment, and the fingerprint information of the guest virtual machine includes kernel structure information and system symbol table information of the guest virtual machine operating system.

Further in step 2), the user sets a whitelist program that is authorized to call the service, and transmits the source file containing the key code segment to the cloud in a safe environment, so that the cloud can generate the key code segment in the monitored client virtual machine. The key code segment includes the kernel image code segment, the kernel module code segment, the code segment of the executable program authorized to call the cryptographic operation service, and the dynamic link library code segment related to the executable program authorized to call the cryptographic operation service. The hash library contains the entry point information of these key code segments, the offset of the code segment, the virtual address where the code segment is loaded into memory, and the hash value of the key code segment calculated in page units.

Further in step 3), the described utilization of the nested page table technology supported in the CPU chip carries out the integrity check to the key code segment loaded in the memory in real time, which can be realized based on the CPU chip of Intel EPT and AMD RVI. .

Further in step 3), through the setting of the permission bit in the nested page table supported by the CPU chip, it is guaranteed that the code in the monitored guest virtual machine is checked for integrity before execution, and the specific steps are as follows:

a) Set the executable permission bit of all memory pages loaded by the guest virtual machine to non-executable;

b) If a page fault exception occurs, VMM handles the exception, if the abnormal cause of the page fault is an instruction fetch error, execute step c); if the cause of the page fault exception is a write error, execute step d);

c) Perform an integrity check on the memory that caused the abnormal instruction fetch, and maintain the whitelist process list according to the check result; for the memory page that passes the check (that is, it belongs to the key code segment), set its page permission to executable, Not writable; for the memory page that fails the verification (that is, it does not belong to the key code page or the key code segment that is damaged), set the permission of the memory page where the page fault exception occurs to executable and writable;

d) Set the writable permission bit of the memory page to writable, and set the executable permission bit to non-executable.

Further, when dealing with an abnormal instruction fetch, the locality principle is used to check the integrity of the code segment, and the whitelist process list is maintained according to the check result. The steps are as follows:

a) Check the virtual address of the monitored guest virtual machine that caused the instruction fetch error, if the virtual address is the kernel space address, then execute step b); if the virtual address is the user space address, then execute step g);

b) Check whether the virtual address is a known kernel address range, if it is a known address space, perform step c); otherwise, perform step d);

c) Calculate the hash value of the code segment on the memory page with the fetch error, and compare the hash value with the hash value of the corresponding kernel code page in the hash library. If the comparison result is not the same, execute step d ); if the same, do nothing;

d) Calculate the hash value of the code segment on the memory page, and compare the hash value with the hash value of the code page where the entry points of all kernel modules are located, if the comparison results are the same, execute step e); otherwise, execute step f);

e) Record the address range of the kernel module, so that if a new code page is executed and an unexecutable error is triggered, and the address falls in this range, the corresponding code page can be found for hash verification; then perform the steps g);

f) Clear the list of blank list processes, and mark the state of the monitored guest virtual machine as a suspicious state; then perform step g);

g) Check whether the CR3 register value (that is, the page directory address of the current process) is in the whitelist process list, and if so, execute step h); otherwise, execute step i);

h) Calculate the hash value of the code segment on the memory page that triggers the value error, and compare it with the hash value of the corresponding code page in the hash library. If the comparison result is not the same, execute step j);

i) Calculate the hash value of the code segment on the memory page that triggers the value error, and compare it with the hash value of the code page where the entry points of all authorized access service programs are located. If the comparison results are the same, and the memory The address of the page is the same as the address of the memory page where the entry point is located, then execute step k);

j) delete the CR3 register value (that is, the page directory address of the current process) from the whitelist process list, and execute step i);

k) Insert the CR3 register value into the whitelist process list.

Further in step 4), once the service invocation monitor detects the occurrence of the service invocation event in the VMM, the fingerprint information of the monitored virtual machine is used to extract the memory of the monitored virtual machine, and a high-level semantic acquisition and invocation service is constructed. Process information related to the event, and obtain the page directory address information of the calling service process in the memory area of the calling process, and then check whether the value exists in the whitelist process table. If it exists, the service call is allowed to pass; if it does not exist , the service call is rejected.

Compared with the prior art, the positive effects of the present invention are:

In the present invention, the service invocation monitoring system can not only monitor the invocation of services provided by the cloud service provider through the network, but also monitor the invocation of other services. For example, a cryptographic computing service provided by a third party through a network, a service provided by a virtual device (eg vTPM) in a virtualization platform, and a service provided by using a physical device, etc.

Description of drawings

Fig. 1 is a schematic diagram of a cryptographic computing service invocation monitoring system;

2 is a schematic diagram of a cryptographic operation service invocation monitoring system that provides a cryptographic operation service by KMS service in an embodiment;

Fig. 3 is the process structure description information diagram in the Linux system in the embodiment;

Figure 4 is a flow chart of integrity verification of key code segments;

FIG. 5 is a schematic diagram of address translation using nested page tables.

Detailed ways

In order to make the above objects, features and advantages of the present invention more clearly understood, the present invention will be further described below through specific embodiments and accompanying drawings.

This embodiment is an example of a cryptographic computing service invocation monitoring system in a cloud environment.

FIG. 2 is a schematic diagram of providing a cryptographic computing service invocation monitoring service in a virtualized environment by adopting the method of the present invention. The user subscribes to the key management service (KMS) provided by the cloud service provider, and the KMS service provides services through https. The KMS service is responsible for managing the user's key and providing cryptographic computing services.

The service call monitoring system in this embodiment is implemented in the QEMU-KVM virtualization platform, and the host computer uses an Intel CPU chip. The detection part of the cryptographic operation service invocation event is implemented in QEMU, and the part involving the use of Intel EPT technology to check the integrity of key code segments is implemented in the KVM kernel module. KVM is a module of the Linux kernel. It implements the function of guest virtual machine acceleration based on hardware virtualization technology. The maintenance of the EPT page table is also responsible for the KVM kernel module. All VM-EXITs are processed by KVM first, and I/O events that cannot be processed by KVM are handed over to QEMU in user space for processing. The guest virtual machine operating system uses the Linux operating system, and the kernel version is 3.13.7.

In the embodiment, the Volatility tool is used to generate a profile of the kernel structure and system symbol information of the monitored guest virtual machine in a safe environment.

In the embodiment, the user sets the program authorized to call the KMS service as A, and the dynamic link library related to A is libc.so.6. Analyze the ELF file of the kernel image, program A and libc.so.6 in a safe environment, take out the load address of the code segment and the information of the code segment, and hash the code segment in page units, Write the hash and the corresponding load address into the hash library in a one-to-one correspondence. At the same time, collect the .ko files in the monitored guest virtual machine, simulate the process of kernel loading and linking .ko, and calculate the offset of the Init segment of the kernel module and the code segment in the Core segment corresponding to each .ko loaded into the memory. The address information, and the corresponding code segment is hashed in units of pages, and the offset address information and the hash are written into the hash library in a one-to-one correspondence.

In the embodiment, starting when the guest virtual machine is started, the permission bits of the entries in the EPT page table corresponding to the physical pages in the loaded memory are set to ensure that all codes loaded into the memory trigger integrity detection before execution. ,Specific steps are as follows:

a) Starting from when the guest virtual machine is started, initialize a whitelist process linked list, and set the page table entry permission bits corresponding to all newly loaded memory pages to non-executable on the EPT page table;

b) When the EPT Violation exception occurs and the guest virtual machine exits, read the error_code in the KVM, if the error_code is an instruction fetch error, execute step c); if the error_code is a write error, execute step d);

c) Integrity check of the code page;

d) Set the permission of the memory page to be writable and non-executable.

In the embodiment, when "EPT violation" is an instruction fetch error, read the value of the CR3 register (the page directory address of the current process) and the value of the CR2 register (the monitored client that caused the instruction fetch error) in the monitored guest virtual machine. The virtual address of the virtual machine) and the content of the code page that caused the instruction fetch error, and then start the integrity check of the code page, as shown in Figure 4, the check steps are as follows:

a) Check the virtual address of the monitored guest virtual machine that caused the error, and if the virtual address is a kernel space address, execute step b); if the virtual address is a user space address, execute step g);

b) Check whether the virtual address is a known kernel address range, if it is a known address space, perform step c); otherwise, perform step d);

c) Calculate the hash value of the code segment on the memory page, compare the hash value with the hash value of the corresponding kernel code page in the hash library, and execute step d) if the comparison result is not the same;

d) Calculate the hash value of the code segment on the memory page, and compare the hash value with the hash value of the code page where the entry points of all kernel modules are located, if the comparison results are the same, execute step e); otherwise, execute step f);

e) record the address range of the kernel module;

f) Clear the blank list process list, and mark the status of the monitored guest virtual machine as suspicious;

g) Check whether the CR3 register value is in the whitelist process list, if so, execute step h); otherwise, execute step i);

h) Calculate the hash value of the code segment on the memory page, and compare it with the hash value of the corresponding code page in the hash library, if the comparison result is not the same, execute step j);

i) Calculate the hash value of the code segment on the memory page, and compare it with the hash value of the code page where the entry points of all authorized programs are located, if the comparison results are the same, then perform step k);

j) delete the CR3 register value from the whitelist process list, and perform step i);

k) Insert the CR3 register value into the whitelist process list.

In the embodiment, a set of ioctl is implemented in the KVM module for QEMU to call, and the user detects whether the page directory address of the current process is in the whitelist process list. When a network packet filtering module detects a network data packet related to KMS, it uses the method of memory analysis to traverse the process linked list and analyze the process that calls the password service, that is, the file handle opened by the process is a socket connection, and the socket connection The address is the IP and port of KMS, and take out the page directory address of the process (the linux process structure is shown in Figure 3), and check whether the page directory address is in the whitelist process table maintained by KVM. If the value is in the whitelist process table, allow the network packet to pass; otherwise, discard the network packet. At the same time, when the data packet returned by KMS is detected, it is checked again whether the page directory address of the calling process is still in the whitelist process list. If it exists, the returned data packet is allowed to enter the monitored guest virtual machine; otherwise, the data packet is discarded.

The above embodiments are only used to illustrate the technical solutions of the present invention rather than limit them. Those of ordinary skill in the art can modify or equivalently replace the technical solutions of the present invention without departing from the spirit and scope of the present invention. The scope of protection shall be subject to what is stated in the claims.

Claims (9)

1. A service call monitoring method in a virtualization environment comprises the following steps:

1) the cloud generates a hash library of key code segments in the monitored client virtual machine according to the authorized access service program set by the tenant and the related source file, wherein the hash library is used for the service call monitor to check the integrity of the key code segments;

2) after the monitored client virtual machine is started, the service call monitor sets a memory page permission limit loaded by the monitored client virtual machine by using a nested page table technology, so that before codes on memory pages are executed, the integrity of key code segments in the memory is verified, a white list process linked list is maintained, and page directory addresses of the white list process are stored in the white list process linked list; the white list process is an authorized process, and the integrity of the code segment is verified;

3) when the service calling monitor detects that the service calling happens, calling a virtual machine state detection module, acquiring a page directory address of a calling process according to the feature of the service calling, and allowing the service calling to be executed when the page directory address of the calling process is in the white list process linked list;

4) and for service call with a return result, the service call monitor detects the return result event of the service call, when the return result event of the service call is detected to occur, the virtual machine state detection module is called, the information of the call process is obtained according to the characteristics of the return result event, and when the call process is in a white list process chain table, the return result of the service call is allowed to pass.

2. The method of claim 1, wherein the checking the integrity of the critical code sections in memory is performed by:

21) setting the executable permission limit of all memory pages loaded by the client virtual machine as non-executable;

22) if a page fault exception occurs and the page fault exception is an instruction fetch fault, executing step 23); if a page fault exception occurs and the page fault exception is a write fault, go to step 24);

23) carrying out integrity verification on the memory causing the abnormal instruction fetching, and maintaining a white list process list according to a verification result; for the memory page passing the verification, setting the page authority of the memory page to be executable and non-writable; for the memory pages which fail to be checked, setting the authority of the memory pages with the abnormal page errors as executable and writable;

24) the method comprises the steps of setting the writable permission limit of a memory page with a page fault exception to be writable, and setting the executable permission limit to be non-executable.

3. The method according to claim 2, wherein in step 23), the method for performing integrity check on the memory page with the value error and maintaining the white list process list according to the check result includes:

a) checking the virtual address of the monitored client virtual machine causing the instruction fetch error, and if the virtual address is the kernel space address, executing the step b); if the virtual address is the user space address, executing step g);

b) checking whether the virtual address is a known kernel address interval, and if the virtual address is a known kernel address space, executing the step c); otherwise, executing step d);

c) calculating the hash value of the code segment on the memory page with the wrong instruction, comparing the hash value with the corresponding hash value in the hash library, and executing the step d) if the comparison result is different;

d) calculating the hash value of the code segment on the memory page, comparing the hash value with the hash values of the code pages where the entry points of all the kernel modules are located, and executing the step e) if the comparison results are the same; otherwise, executing step f);

e) recording the address interval of the kernel module;

f) clearing a white list process list, and marking the state of the monitored client virtual machine as a suspicious state;

g) checking whether the page directory address of the current process is in the white list process linked list, if yes, executing the step h); otherwise, executing step i);

h) calculating the hash value of the code segment on the memory page with the triggered value error, comparing the hash value with the hash value of the corresponding code page in the hash library, and executing the step j) if the comparison result is different;

i) calculating the hash value of the code segment on the memory page with the triggered value error, comparing the hash value with the hash values of the code pages where the entry points authorized to access the service program are located, and if the hash values are the same and the addresses of the memory page and the memory page where the entry points are located are the same, executing step k);

j) deleting the page directory address of the current process from the white list process list, and executing the step i);

k) and inserting the page directory address of the current process into the white list process list.

4. The method according to claim 1, wherein when the service call monitor detects the occurrence of a service call in the VMM, extracting a memory of the monitored virtual machine according to fingerprint information of the monitored virtual machine, constructing a high-level semantic to acquire process information related to the call service, and acquiring page directory address information of the calling service process in a memory area of the calling process, and then checking whether the page directory address exists in the whitelist process table, and if so, allowing the service call to be executed; if not, the service call is rejected.

5. The method as claimed in claim 1, wherein the hash library comprises entry point information of the critical code section, an offset of the critical code section, a virtual address of the critical code section loaded into the memory, and a hash value calculated in units of pages in the critical code section.

6. The method of claim 1 or 5, wherein the critical code sections include a kernel image code section, a kernel module code section, a code section of an executable program authorized to access the service, and a code section of a dynamically linked library associated with the executable program authorized to access the service.

7. A service call monitoring system in a virtualization environment is characterized by comprising a hash library and a service call monitor; wherein,

the service call monitor is responsible for setting a memory page permission limit loaded by a monitored client virtual machine by using a nested page table technology when the monitored client virtual machine is started, checking the integrity of key code segments in a memory in real time before executing codes on the memory page, and maintaining a white list process linked list, wherein the white list process linked list stores page directory addresses of white list processes; the white list process is an authorized process, and the integrity of the code segment is verified; when detecting that the service call occurs, calling a virtual machine state detection module to acquire a page directory address of a calling process, and allowing the service call to be executed when the page directory address of the calling process is in the white list process list; for service call with a return result, the service call monitor detects a return result event of the service call, when the return result event of the service call is detected, the virtual machine state detection module is called, information of a call process is obtained according to the characteristics of the return result event, and when the call process is in a white list process list, the return result of the service call is allowed to pass;

and the hash library is used for storing the hash of the key code segment in the monitored client virtual machine, so that the service call monitor can check the integrity of the key code segment.

8. The system as claimed in claim 7, wherein the hash library includes entry point information of the critical code section, an offset of the critical code section, a virtual address of the critical code section loaded into the memory, and a hash value calculated in units of pages in the critical code section.

9. The system of claim 7 wherein the critical code sections include a kernel image code section, a kernel module code section, a code section of an executable program authorized to access the service, and a code section of a dynamically linked library associated with the executable program authorized to access the service.