US20060259950A1

US20060259950A1 - Multi-layer system for privacy enforcement and monitoring of suspicious data access behavior

Info

Publication number: US20060259950A1
Application number: US11/357,741
Authority: US
Inventors: Ulf Mattsson
Original assignee: Protegrity Corp
Current assignee: Protegrity Corp
Priority date: 2005-02-18
Filing date: 2006-02-17
Publication date: 2006-11-16
Also published as: WO2006089277A2; US8701191B2; GB0716647D0; US20090025057A1; GB2438133A; US20140165202A1; US20150096049A1; KR20070114725A; US20130174215A1; US8935787B2; WO2006089277A3

Abstract

A method for controlling data access in a data-at-rest system includes executing a link intrusion prevention analysis between multiple layers of the data-at-rest system, introducing a privacy policy at enforcement points that span multiple system layers, and dynamically altering the privacy policy.

Description

RELATED APPLICATION

This application claims priority from co-pending provisional U.S. Application Ser. No. 60/654,181, filed Feb. 18, 2005. This application also claims priority from copending U.S. application Ser. No. 10/034,996, filed Dec. 28, 2001.

FIELD OF DISCLOSURE

The disclosure is directed to software for interacting with a database, and in particular, to software for interacting with databases that include encrypted data.

BACKGROUND

It is difficult to detect advanced attacks on data and data misuse by monitoring only one system layer. It is likewise difficult to attain acceptable performance in using external policy driven encryption systems.
One difficulty arises because large amounts of encrypted data are exposed when providing an efficient search on encrypted data. In addition, large amounts of sensitive data are exposed when using effective performance optimization to offload cryptographic operations. This results in exposure of data in memory or disk, outside of the control of the security/encryption system.

SUMMARY

In general, in some aspects, a method for controlling data access in a data-at-rest system includes executing a link intrusion prevention analysis between multiple layers of the data-at-rest system, introducing a privacy policy at enforcement points that span multiple system layers, and dynamically altering the privacy policy.
In some implementations, the method includes one or more of the following features. The data-at-rest system is a database system. The method further includes modifying the protection of data at one of the multiple system layers. The step of modifying is performed based on a result of the link intrusion prevention analysis. The privacy policy includes access control information. The privacy policy includes intrusion detection information. The privacy policy includes cryptographic information.
In general, in some aspects, a method for controlling access to a database system includes assigning a first access criterion and a second access criterion to a user role, receiving a query from a user, the user having an access history, determining that the user matches the user role, comparing, in a first system layer, the access history to the first access criterion, and comparing, in a second system layer that differs from the first system layer, the access history to the second access criterion.
In some implementations, the method includes one or more of the following features. The first access criterion comprises a privacy policy. The method further includes learning a value for the first access criterion. The method further includes selecting a response to the query, wherein the response is selected from the group consisting of blocking the query, alerting a system administrator and allowing the query, and allowing the query. Selecting a response to the query comprises selecting a response to the query based on a result of the step of comparing in a first system layer.
In general, in some aspects, a method for accessing data includes in a first system layer, receiving a first request from a user, the user having an access history, the access history including a counter, in the first system layer, comparing the counter to a first threshold, transmitting a second request to a second system layer, the second request being based on the first request.
In some implementations, the method includes one or more of the following features. The method further includes comparing the counter to a second threshold. The counter includes a scorecard. The method further includes determining that the counter exceeds a third threshold, and alerting a system administrator. The method further includes, in the first system layer, transmitting a notification to the second system layer to deny the second request.
Other general aspects include other combinations of the aspects and features described above and other aspects and features expressed as methods, apparatus, systems, program products, and in other ways.
Advantages and features will become apparent from the following description and claims.

DESCRIPTION OF DRAWINGS

FIGS. 1, 3 and 6 are block diagrams of database systems.
FIGS. 2A, 2B, 2C, 4A, 4B, and 5 are flow charts.

DETAILED DESCRIPTION

The system described herein is intended to be integrated into that described in copending U.S. application Ser. No. 10/034,996, filed Dec. 28, 2001, and entitled METHOD FOR INTRUSION DETECTION IN A DATABASE SYSTEM, the contents of which are herein incorporated by reference.
A method and system for overcoming the foregoing difficulties provides for the introduction of a privacy policy with enforcement points that span multiple system layers.
The privacy policy is coupled with link intrusion prevention analysis between multiple system layers. The scope, both in data and in time, for enforcing data privacy and encryption is then dynamically optimized between multiple system layers.
As used herein, multiple system layers includes application database sessions, table data access, table space access, and database file level access. The term “transaction” is intended to include queries. The term “data at rest” is intended to include all forms of stored data. A “data-at-rest system” includes any system for storing data.
In a system for overcoming the foregoing difficulties, selected rules control the amount of data that is exposed, and the time window for exposure of unencrypted data. A policy underlying the selected rules defines the extent to which data privacy is to be enforced for particular data. This extent, which includes the extent of the particular data exposed and the duration of such exposure, is determined on the basis of the sensitivity of the particular data.
Dynamic control over the extent and duration of unencrypted-data exposure required to satisfy a user transaction is provided by linking the intrusion detection point (“IDP”), the policy enforcement point (“PEP”), the audit generation point (“AGP”), and the data-at-rest encryption point (“DEP”). These scopes are controlled by an operational sensitivity class defined in the policy. The operational sensitivity class defines what rules to check and when to do so by linking the IDP, the PEP, the AGP, and the DEP.
At the intrusion detection point, a scorecard is provided to accumulate violation attempts. On the basis of the number of violation attempts, session statistics, and data access statistics spanning multiple system layers, one can determine whether a threshold indicative of an attack has been reached.
A system as described above enhances the ability to detect advanced attacks on data as well as instances of data misuse. The system also reduces the extent to which data is exposed and outside the control of the security/encryption system, both in terms of the amount of data being exposed and the duration of such exposure. In addition, the system enables effective performance optimization and offloading of cryptographic operations.
In an exemplary security system 116, depicted in FIG. 1, a user 115 communicates through a client 114, which interacts with an application server 113. The application server 113 communicates with a PEP 101 to request authorization and to transmit auditing data. The application server 113 also passes queries along to a database 107, which itself communicates with the PEP 101 to request authorization. The database process 107 also communicates with a DEP 103 to request decryption. The DEP 103, in turn, utilizes a hardware security module (“HSM”) 105 and a software security module (“SSM”) 106.
The database process 107 transmits requests to its buffer 110, which (through a process overseeing the buffer) sends audit information to the PEP 101. The buffer process then transmits requests to a file system file 108, which also communicates with the DEP 103 to request that a file be decrypted.
The PEP 101 communicates directly with the DEP 103 to provide authorization for other components to decrypt files. The PEP 101 interacts with the AGP 104 to store audit information. The PEP 101 calls the IDP 102 to provide information used in determining whether a given query should be allowed. In response, the IDP 102 tells the PEP 101 that a given query should either be allowed, blocked, or allowed but with an alert sent to the system administrator. It performs this task with the aid of intrusion detection rules 112 and a scorecard 111 associated with each user.
The DEP 103 can be optimized to perform at a layer that allows granularity (e.g., operations on a table cell vs. a table vs. an entire database vs. an entire file system) in compliance with a privacy policy. The DEP 103 can then dynamically dispatch an operation to be performed, either by a hardware security module (“HSM”) 105 or by a software encryption engine 106, or a combination thereof. The DEP 103 can operate on an in-memory database or on a disk.
Operations on different levels of granularity may be achieved, in the depicted example, by associating the DEP 103 with multiple layers of the database hierarchy. The DEP 103 is connected to The database process 107 and the data store (file system) layer 108. An encryption request originating at the database layer 302 permits the DEP 103 to encrypt data in an individual row, column, or cell. (It might also, however, permit a database administrator to decrypt data for which the administrator lacks authorization.) In some embodiments, an encryption request originating at the file system layer 108 permits the DEP 103 to encrypt data in an individual file system file, thereby preventing a database administrator from accessing sensitive data.
If permitted by the privacy policy, the DEP 103 can, under certain conditions, dynamically re-route a decryption request from a software security module 106 to a hardware security module 105. Exemplary conditions include having a message size larger than a predetermined size. This dynamic re-routing optimizes performance and offloads cryptographic operations.
Upon detecting an attack, the PEP 101 can carry out any combination of the following options: issuing a security alert, blocking access to selected data, disabling one or more users, and disabling a request.
FIGS. 1 and 2A-2C illustrate one example of the operation of a system including a PEP 101, an IDP 102, a DEP 103, and an AGP 104. In this example, a database user 115 initiates a transaction through a client 114, such as a web browser (step 201). The client 114 sends a request to an application server 113, e.g., a web server (step 202).
The application server 113 initiates an authentication request with the PEP 101 (step 203). The PEP 101, in conjunction with the IDP 102, verifies the user's authorization, as described in more detail below in connection with FIGS. 5A and 5B (step 204). If the user is not authorized (step 205), the PEP blocks the query (step 213). If the user is authorized, then the application server 113 sends auditing information to the PEP 101 (step 206), which the PEP 101 transmits to the AGP 104 (step 207). Audit information includes the database user ID, the date and time, the SQL query and other action details, the originating machine name or IP address, and the database name.
The application server 113 then sends a request to a database 107 (step 208). The database process 107 again seeks authorization from the PEP 101 (step 209). The PEP 101 again in conjunction with the IDP 102 verifies the user's authorization as described in connection with FIGS. 5A and 5B (step 210). If authorization is granted (step 211), the PEP 101 transmits the authorization to the database process 107 as well as to the DEP 103 (step 212). The authorization to the DEP 103 indicates that The database process 107 is permitted to access decryption keys associated with columns to which the user 115 has access. If authorization is refused, the PEP 101 blocks the query (step 213) and The database process 107 returns an error (step 214), which the client 114 propagates to the user 115 (step 215).
If the PEP 101 grants authorization, a database process 107 accesses a file 108 in the file system, through the database's buffer 110 to read the relevant data (step 216). A computer process overseeing the buffer 110 sends additional audit information to the PEP 101 (step 217), which the PEP 101 transmits to the AGP 104 (step 218). If the database file 108 is encrypted (step 219), the file system requests that the DEP 103 decrypt the file (step 220). If the DEP 103 has been previously authorized by the PEP 101 in step 212 (step 221), then the DEP 103 decrypts the file using a hardware security module (“HSM”) 105 and/or a software decryption engine 106 (step 222), and returns the requested contents (step 223).
The database process 107 then checks to see if any of the requested information is in an encrypted column (step 224). If so, the process overseeing the database process 107 requests that the DEP 103 decrypt the relevant columns (step 225). The DEP performs the requested decryption using the HSM 105 and/or the software decryption engine 106 (step 226), and returns the decrypted results (step 227).
The database process 107 extracts the relevant information (step 228) and returns it to the application server 113 (step 229). The application server 113 returns a result to the client 114 (step 230), which displays a result to the user 115 (step 231).
Often, many application users will share a single database user. In some examples, a PEP 101 connected to the database server utilizes the identity of the application user (in addition to or instead of the database user) as a factor in determining whether a given request is authorized. The PEP 101 does this by communicating with an application user mapping table located within the application server's security system 312. The table contains a mapping associating the application user with the database user. Real time mapping data provides information about which application user is using the database connection at any given time. In some examples, the mapping table is stored in a database table. In other examples, the mapping table is stored in a file. In still other examples, the table, or simply the identity of the application user, is transmitted by the application server to the database server during the session.
As illustrated in FIG. 3, in some examples, security components are connected to three levels: the web or application server 301; the database 302; and the data store or file system 303. Services on these three levels communicate across multiple channels: the data request channel 304; the session information channel 305; and the directory information channel 306. In the depicted example, user A 309 is logged in to an application. The application requests data, as user B 310, over the data request channel 304, from the database. The database is running as user C 311 on a server, and requests data, over another data request channel, from the data store (e.g., the file system).
Meanwhile, on the application server 301, a mapping table, which associates user A 309 (the application user) with user B 310 (the database user), is maintained. This information may be communicated, via a separate session information channel 305, to the database's security system 307.
Examples of further details of the operation of the PEP 101, the IDP 102, and the DEP 103 are provided below. The boundaries of the functions performed by the IDP 102, the PEP 101, the AGP 104, and the DEP 103 are not fixed; some functions may be combined in a single component, or allocated differently between components.
A. PEP
FIG. 4A explains in further detail an example of operation of the PEP 101 (see FIG. 2, steps 204 and 210). First, the PEP 101 receives a request for authorization (step 401). The PEP 101 then retrieves the user's identity and corresponding group, as described below in connection with FIG. 4B (step 402). The PEP 101 then determines whether the user or group is authorized to access the requested data, for example, by consulting a privacy policy or access control list (step 403). If the user or group is unauthorized to access the requested data, the PEP 101 skips to step 410.
If the user or group is authorized, the PEP 101 then retrieves session variables (including the time of day, day of week, IP address from which the user is logged in, the user's geographic location, the user's identity, the user's group, the user's client's software, etc.), and stores these variables as the user's “role” (step 404). The PEP 101 then communicates with the IDP 102 to determine whether the query is valid for the user's role (step 405). The IDP makes this determination in a process exemplified by FIG. 5. The PEP 101 determines whether the IDP 102 allowed the query (step 406). If the IDP 102 rejects the query, the PEP 101 skips to step 410.
If the query was allowed, the PEP 101 checks to see whether the IDP 102 indicated that an alert was to be sent to the system administrator (step 407). If so, the PEP alerts the administrator (step 408). In either event, the PEP 101 authorizes the query (step 409). If the query was not allowed, the PEP 101 denies authorization for the query (step 410).
FIG. 4B describes how, in some examples, the database's security system 307 retrieves the user's identity and corresponding group in step 402. First, the PEP 101 in the database's security system 307 opens a session information channel 305 to the application server 301 (step 451). Next, the PEP 101 requests, from a mapping table in the application's security system 312, the application user corresponding to the current database user (for example, user B 310) (step 452). The application server responds, for example, that the corresponding user is user A 309 (step 453).
Next, the PEP 101 opens a separate directory information channel 306 back to the application server 301 (step 454). On the directory information channel 306, the PEP 101 requests the group mapping for user A 309 (step 455). In a typical response, the application server 301 indicates that user A 309 is a member of group X (step 456).
FIG. 4B describes the process by which the PEP 101 in the database's security system 307 retrieves information from the application server 301. In some examples, the PEP 101 in the data store's security system 308 requests the identity of the database user (i.e., user B 310) from the PEP 101 in the database security system 307 over the session information channel 305. The PEP 101 in the data store's security system 308 may also ascertain the database user's group over the directory information channel 306.
In some examples, the PEP 101 in the data store's security system 308 can identify the application user by requesting the information from the PEP 101 in the database's security system 307, which then relays the query to the application server's mapping table 311. Similarly, in some examples, the PEP 101 in the data store's security system 308 can ascertain the application user's group, by requesting the information from the PEP 101 in the database's security system 307, which then relays the query to the mapping table 311 in the application server.
In some practices, a database's security system 307 (for example, through its PEP 101) notifies other layers to indicate that a severe attack has occurred. In some practices, the IDP 102 in the application server's security system 312 receives this notification and subsequently blocks all access attempts that would otherwise have only triggered an alert to the system administrator. In some practices, the DEP 103 in the data store's security system 308 receives this notification and blocks all subsequent requests to decrypt data.
One example of these practices is provided where a PEP 101 detects that authorized access to credit card information at the database level exceeds normal usage, but not is not at a critical level. The PEP 101, in this example, modifies a privacy policy to instruct the application server's security system 312 to block further access attempts. In another example, the PEP 101 in the application server's security system 312 detects multiple hacking attempts from multiple locations. The security system 312 modifies a privacy policy to block requests at the application server 312 level, increase file security at the data store's security system 308, and prevent the data store's security system 308 as well as the database's security system 307 from decrypting sensitive data.
B. IDP
In some embodiments, the IDP 102 has a learning mode and an enforcement mode. In learning mode, the IDP 102 acquires information about users of the system, including the typical time of day and day of week during which they access the system, the resources they usually access, their physical location or IP address, and the volume of data they usually access. In some examples, the IDP 102 maintains a Bayesian network to associate authorized accesses with these variables. In other examples, other types of learning may be used. When the IDP 102 is in enforcement mode, it denies access to a user when the time or day of access, the resources accessed, the user's location or IP address, or the volume of data requested exceeds a learned threshold or differs from learned values. The IDP 102 optionally alerts a system administrator when any of these criteria exceeds a learned threshold or differs from learned values.
In some embodiments, the IDP 102 accepts user logins only during certain times of day, or only on certain days of the week, or only from certain physical locations. In some examples, the IDP 102 learns how these criteria should be restricted. In other examples, the system administrator manually enters restrictions. In some examples, the system administrator manually changes restrictions, for example, to temporarily allow a particular user to log in from a distant location when the user is on vacation.
In some embodiments, the IDP 102 restricts the volume of data a user may access in a given day. In one example, the IDP 102 permits a user to access only a predetermined number of rows per day from a given table. In another example, the IDP 102 permits a user to issue only a predetermined number of queries per day in a given table. In other examples, the user is restricted to a given volume of data over the entire database, rather than in specific tables. In some examples, the IDP 102 uses a counter to maintain information about the volume of data a user has accessed by means of a counter.
In some examples, an IDP 102 restricts access based on the user's role. A user's role may be based on his or her identity, the time of day, the day of week, the IP address being used, the country or geographic region from which the request originates, etc. In some examples, an IDP 102 located in the database server sets a maximum number of rows per day accessible to users in a given role. Some examples restrict the number of rows a user in a given role may insert, or the number of rows a user in a given role may modify, or the number of rows a user in a given role may delete. In some examples, these values are learned while the IDP 102 is in learning mode.
Some examples permit the IDP 102 and/or the PEP 101 to communicate with a trusted component running on an authorized client to further assist in user authentication.
In some examples, the IDP 102 utilizes one or more of the following criteria to decide whether to permit access, block access, or alert the administrator: session authorization (i.e., the user's identity); session authentication (i.e., the resources a user is entitled to access); session encryption; password integrity; database software integrity; application data integrity; database metadata integrity; security software integrity; time of day of access; and signature rules (i.e., pattern matching and content analysis to detect any known attack signature using, e.g., Snort® network intrusion detection software). To verify data or software integrity, a hash value is stored and verified against periodically.
In some examples, the IDP 102 triggers an alert whenever a particular user accesses an abnormally high volume of data. When this alert is triggered, the PEP 101 analyzes an audit log to ascertain whether unusual activity is occurring. If so, the IDP 102 can disallow further accesses by the current user, and/or send an alert to a system administrator.
An example of the operation of the IDP 102 will now be described with reference to FIG. 5. First, the IDP 102 receives a request for authorization from the PEP 101 (step 501). The request includes information about the user's role (see FIG. 4A, step 404). The request also includes a query the user seeks to execute. The IDP 102 next retrieves the user's scorecard 111 (including variables tracking a user's total volume of access in a given time period, e.g., total number of rows accessed in a day, total kilobytes of data downloaded in a day, etc.) (step 502).
Next, the IDP 102 checks to see if it is in learning mode (step 503). If it is, then the IDP 102 updates a Bayesian network to learn that the query is authorized for this user role (step 504). Next, the IDP 102 adds data from the current query to the user's scorecard 111 (step 505). Finally, the IDP 102 transmits authorization to the PEP 101 (step 506).
If the IDP is not in learning mode, then the IDP 102 calculates the probability that the query is allowed for the user's role (step 507). If this probability is below a predetermined threshold a (step 508), then the IDP 102 tells the PEP 101 to block the query (step 509). If, instead, the probability is between the threshold a and a predetermined threshold b, where b>a (step 510), then the IDP 102 adds the data from the current query to the scorecard 111 (step 511), allows the query, and tells the PEP 101 to send an alert to a system administrator (step 512). If the probability is greater than the threshold b, the IDP 102 simply allows the query (step 513).
More generally, in some examples, the IDP 102 maintains a set of i thresholds t_i. For each interval [t_i, t_i+1), a different action k_iis defined. If the probability that the query is allowed is within the interval [t_i, t_i+1), then the action k_iis performed.
In some examples, the IDP 102 increments the value on the scorecard 111 in response to accesses, or attempted accesses, to sensitive resources. In some examples, sensitive resources include access to prespecified applications and prespecified network addresses. In other examples, the IDP 102 increments the value of the scorecard by a greater amount in response to a failed or disallowed attempt to access the sensitive resource.
C. DEP
FIG. 7 depicts an example of interaction between a DEP 601 connected to a database 603 and a DEP 602 connected to a file system 604. In one example view 605, two columns are encrypted, but the table itself is not. In this example, authorization is handled entirely by the database's DEP 601. In a second example view 606, one column is encrypted, and the table is encrypted as well. In this example, the database's DEP 601 provides the decryption key for the column, while the file system's DEP 602 provides the decryption key for the table. In this example, a database administrator would be precluded from accessing secure data due to the table-level encryption. In the third example view 607, no columns are individually encrypted; but the table is encrypted. In this example, the file system's DEP 602 provides the decryption key.
In some examples, when a PEP 101 determines that a given query is to be blocked, it performs one or more of the following tasks: disconnecting the user; denying access to cryptographic keys; writing a record in a log file; and sending an error return code, coupled with no data, back to the requesting application.
In some examples, the PEP 101 includes a machine and program authorization (MPA) component. This component prevent or restrict users with valid login names and passwords from connecting to the database unless they access the database from a machine that has been preauthorized. Machines are authorized if they have an authorized IP address, and additionally, if they are able to specify both the port on which the database server is listening and the name of the database.
In some examples, the PEP 101, IDP 102, DEP 103, and AGP 104 are part of the Protegrity™ Secure Data server, which is available from Protegrity Corporation of Stamford, Conn.
It is to be understood that while the invention has been described in conjunction with the detailed description thereof, the foregoing description is intended to illustrate and not limit the scope of the invention, which is defined by the scope of the appended claims.

Claims

1. A method for controlling data access in a data-at-rest system, the method comprising:

executing a link intrusion prevention analysis between multiple layers of the data-at-rest system;

introducing a privacy policy at enforcement points that span multiple system layers; and

dynamically altering the privacy policy.

2. The method of claim 1, wherein the data-at-rest system is a database system.

3. The method of claim 1, further comprising modifying the protection of data at one of the multiple system layers.

4. The method of claim 3, wherein the step of modifying is performed based on a result of the link intrusion prevention analysis.

5. The method of claim 1, wherein the privacy policy comprises access control information.

6. The method of claim 1, wherein the privacy policy comprises intrusion detection information.

7. The method of claim 1, wherein the privacy policy comprises cryptographic information.

8. A computer-readable medium containing instructions for causing a computer to perform the method of claim 1.

9. A method for controlling access to a database system, the method comprising:

assigning a first access criterion and a second access criterion to a user role;

receiving a query from a user, the user having an access history;

determining that the user matches the user role;

comparing, in a first system layer, the access history to the first access criterion; and

comparing, in a second system layer that differs from the first system layer, the access history to the second access criterion.

10. The method of claim 9, wherein the first access criterion comprises a privacy policy.

11. The method of claim 9, further comprising learning a value for the first access criterion.

12. The method of claim 9, further comprising selecting a response to the query,

wherein the response is selected from the group consisting of blocking the query, alerting a system administrator and allowing the query, and allowing the query.

13. The method of claim 12, wherein selecting a response to the query comprises selecting a response to the query based on a result of the step of comparing in a first system layer.

14. A computer-readable medium containing instructions for causing a computer to perform the method of claim 9.

15. A method for accessing data, the method comprising:

in a first system layer, receiving a first request from a user, the user having an access history, the access history including a counter;

in the first system layer, comparing the counter to a first threshold; and

transmitting a second request to a second system layer, the second request being based on the first request.

16. The method of claim 14, further comprising comparing the counter to a second threshold.

17. The method of claim 14, wherein the counter comprises a scorecard.

18. The method of claim 14, further comprising:

determining that the counter exceeds a third threshold; and

alerting a system administrator.

19. The method of claim 14, further comprising:

in the first system layer, transmitting a notification to the second system layer to deny the second request.

20. A system comprising:

a computer-readable medium as recited in claim 14; and

a computer in data communication with the computer-readable medium.