Patent Application
20060085856
The present application claims priority to European Patent application No. 04292461.3, filed Oct. 15, 2004.
The present invention relates generally to computer systems security and more particularly to a system and method of detecting invalid attempts to log on to such computer systems to access a computer network.
In today's wide network enterprise, security has become a major concern to prevent unauthorized access to the many computer systems of the network. Computer systems within one business site are typically connected using a Local Area Network (LAN) and a Network Administrator is responsible for keeping the network up and running properly. As local area networks (LANs) continue to proliferate, and the number of personal computers (PCs) connected to LANs continue to grow at a rapid pace, network security becomes an ever increasing problem for network administrators. As the trend of deploying distributed LANs continues, this provides multiple access points to an enterprise network. Each of these distributed access points, if not controlled, is a potential security risk to the network. Among these risks, a virus attacks strongly impacts all IT infrastructures by the very fast spreading of the virus. A specific kind of virus is the well-known ‘Worm’; one which makes use of security loopholes in operating systems and spreads from one system to another via networks through the standard TCP/IP ports 137, 139 and 445. Another risk is the volunteer attack and the unauthorized access to protected resources (also using spoofing).
Log files and other accounting mechanisms can be used to track users and their activities. Using log files and audit information is known as passive detection since these rely on a passive analysis set of data. The system administrators are responsible for reviewing the operating system security event logs to determine if a system attack or breach of security has occurred. Some known products allow administrators to review those logs, such as Tivoli Risk Manager from the Assignee and GFi LANguard S.E.L.M. from GFI Software Ltd.
Tivoli Risk Manager is based on a framework infrastructure allowing cross operating system availability (OS/2, Windows, Linux and AIX) to report the alerts on a centralized console. This solution offers the possibility for the administrator to create the appropriate relationships between various security alerts. However, the relationships concerning the Netbios invalid attempts between different operating systems are not deeply analyzed and do not relate to external databases.
GFi LANguard S.E.L.M. is a security event log monitor that collects all security events in one central database, creates reports and custom filters. This solution suffers from the limitation of operating on one Operating System only, namely Windows, thereby leaving attacks possible for others operating systems.
More generally, all existing solutions leave administrators with receiving a large number of false warning messages that compel spending a lot of time manually analyzing those messages and sorting out the non relevant violations from the true ones.
Therefore, there is a need for a solution that overcomes the aforementioned drawbacks. The present invention offers such solution.
Accordingly, the main object of the invention is to provide a system and achieve a method for improving computer system security by determining computer system invalid logon attempts. Particularly, the invalid logon attempts contain the Windows and the OS/2 Netbios bad logons (on TCP ports 137, 139 and 445), the Linux bad logons while accessing Samba resources, and the OS AIX logons.
It is another object of the invention to provide a unique central database to analyze the invalid logon attempts.
Another object of the invention is to provide a central database to manage any operating system violations.
A further object of the invention is a to offer a solution to easily and efficiently identify in an enterprise computer network the machines infected by WORM viruses.
Yet another object of the invention is to provide a solution involving less human resources to analyze computer system violation events thereby allowing fast proactive reaction.
Other objects of the invention will in part be obvious, and in part appear hereinafter.
According to an aspect of the present invention, a system is proposed for detecting the invalid access of a computer device to a computer network.
More generally, the method associated with the present invention preferably operates in a computer network having computer servers operating on different operating systems and a plurality of computer devices. Each computer device is managed by a computer server at the operating system level. The computer network includes a plurality of information databases that contain information associated with the users and with the computer devices of the computer network. The method generates on each computer server a set of identifying files for each computer device managed by the computer server. All sets of identifying files from the plurality of computer servers are then gathered into a unique central violation database. Links are created between each set of identifying files and the plurality of information databases in order to determine a level of network access violation for each computer device. Violation messages for each computer device are generated based on the level of network access violation.
In a commercial form, computer readable program means allowing a computer machine to perform the preferred method is embodied on a program storage device that is readable by the computer machine.
The above and other objects, features and advantages of the invention will be better understood by reading the following Detailed Description of the invention in conjunction with the accompanying drawings wherein:
FIGS. 3-a to 3-c is a flow chart of the main steps of the process of the invention run at the central analyzing machine.
Before referring to the figures, some general considerations as to the terms used and as to the advantages of the present invention are first highlighted:
A central analyzing tool (also named hereinafter central LogLooker) runs on a central analyzing machine. When run, the tool analyzes all the violations in one central database (named DBLOG hereinafter) that is continuously updated with information received from O/S servers. A local analyzing tool (also named hereinafter local LogLooker) is installed on each O/S server to run with the corresponding Operating System (OS/2, Windows, Linux or AIX).
Originator devices that create violation information into the local LogLookers are named attacker devices. Those attackers devices are identified inside the logs by their workstation names (or machine name referred to as <workstation> in the text).
The method of the present invention offers the advantage of analyzing the relationships existing between violations occurring on different local LogLookers. Moreover, the analysis of the violations may be advantageously linked to different databases, either internal to a company (e.g. Company Employee database) or external (e.g. list of authorized workstation names, list of authorized users, etc . . .).
Finally, it is to be appreciated that the present invention offers a global monitoring of system violations.
Different levels of warning are raised for the violations depending on the analysis of several criteria:
According to the result obtained by checking those different criteria, a level of violation is raised that may be:
Referring first to
A local Loglooker is installed on each managed O/S server (22,24,26,28) which runs it regularly to send automatically the results files (sets of identifying files) to the central analyzing machine 30. It is to be noted that the frequency of the local LogLooker execution may be parametred for each O/S server.
A central Loglooker is installed on the central analyzing machine 30. This latter may be any computing device connected to a network entry point that is permitted to receive the FTP protocol. The central LogLooker runs in loop mode to systematically execute the following tasks:
Thereafter, the process resumes automatically.
Referring now to
Step 201: The execution of the local LogLooker is scheduled via the Windows Task Scheduler.
Step 202: Several commonly used transfer parameters are extracted from the O/S server file (specifically the ‘CONFIG.DBF’ file), and particularly:
Step 203: The current size of the Security Event log file is compared to the maximum size; and
The structure of the ‘LOG’ file is exemplified below:
Step 204: If the maximum size is reached, the content of the current Security Event Log file is saved with a unique name referring to the timing and location of the file (preferably in the form of ‘<workstation+date+month+day+hour+seconds>.evt’). Then the current file is cleaned.
Step 204 offers the following two advantages:
Log Reliability:
Managing the size of the current Security Event Log file ensures a full availability of the log and avoids the saturation of the file before the retention time is not reached.
Performances:
The cleaning operation of the Security Event file ensures a low impact on the CPU of the O/S servers and allow a very short duration of the execution.
Step 205: The local LogLooker creates a batch file which is generated at each execution of the local LogLooker procedure to execute next steps 206 to 207 in batch mode.
Step 206: An empty text file is created and generated at each execution of the local LogLooker procedure (preferably named <workstation.FLAG>. The value of the variable ‘workstation’ is automatically retrieved from the workstation name for the Operating system. As it will be fully explained below (on step 209), this empty file will be used by the central LogLooker.
Step 207: The local LogLooker extracts the information relative to the local users from the Operating System and creates a local userid table (preferably named ‘<workstation>.DBF’ where ‘workstation’ refers as above to the workstation name for this O/S server.
The structure of this table is illustrated below:
Step 208: a file containing the data that the FTP automatic transfer needs to run is created. Preferably the file is a text file including the central analyzing machine address, the userid and password for the logon FTP and the sending commands. This file is generated at each execution of the local LogLooker.
Step 209: the Security Event log file is read. All failed logons (considered as violations) and all revoked events are written into a local violation file (preferably named ‘<workstation>.txt’ where ‘workstation’ refers to the workstation name for the O/S server).
Step 210: the files previously created on steps 204, 206, 207 and 209 are sent to the central analyzing machine via the FTP protocol. Preferably, the files are sent in the following order:
The ‘<workstation>.txt’ file containing the bad-logons information (the violations);
The empty file is sent in last to ensure completion of the transfer of the other files.
The person skilled in the art will readily appreciate that the process of the local LogLooker is easily adapted for the Operating System supported by respective O/S server.
For example, for OS/2 servers, in a preferred implementation, the execution processes as follows:
The execution of the local LogLooker is scheduled via the well-known built-in ‘TIMEXEC’ function of the OS/2 Servers.
The other steps are identical to the process of
At the end of the process, the created files are sent to the central LogLooker via the FTP protocol in the following preferred order:
For the Linux and the AIX Operating Systems, except the scheduling which is executed via the well-known ‘crontab’ command, all steps remain identical to the Windows procedure.
Referring now to
Step 301: all configuration parameters are read:
Step 302: all OS/2 ‘<workstation>.BKP’ files having a correspondant ‘<workstation>.FLAG’ file are read from the FTP incoming directory in order to avoid any uncompleted transferred file. This allows the local LogLookers to send their data without synchronization with the cycle of the central LogLooker.
Step 303: all read files are stored in a dedicated OS/2 memory area of the central analyzing machine.
Step 304: The process verifies the end of file of the memory array. If the end is reached, next step is 307; otherwise the process loops to steps 305 and 306.
Step 305: Each <workstation.BKP>file is renamed (by adding the month, date, hours, minutes and seconds) to declare one unique name for each file.
Step 306: All renamed ‘.BKP’ files are moved from the incoming directory to an archive directory.
This is an important feature of the present invention to have all results and saved Event Log files archived. This ensures that, in case of a malicious update by a hacker of the original log file, a saved copy is available in a secure machine.
From step 307 to step 311: similar operations as for previous steps 302 to 306 are executed but for all ‘*.EVT’ files issued from O/S servers operating on Windows.
Step 312: The ‘*.TXT’ files are imported into a temporary violation database. This step collates into a same database the information extracted from different operating systems.
Step 313: Each violation is compared with the last recorded information existing in the central violation database ‘DBLOG’.
Step 314: If the violation is newer than the content of the ‘DBLOG.DBF’ database for this local LogLooker and this userid, then the new record is added to the central database. Otherwise, the process loops to step 312.
The ‘DBLOG’ database is the violation database that contains all violations for all local LogLookers. Below is a figure to illustrate the structure of this database.
Step 315: Each ‘<workstation>.TXT’ file is renamed (by adding the month, date, hours, minutes and seconds) to have one unique name. Each renamed file is moved from the incoming directory to the archive directory. After this step, the ‘<workstation>.FLAG’ file is deleted.
Step 316: all ‘<workstation>.DBF’ files having a correspondant ‘<workstation>.FLAG’ file are read from the FTP incoming directory to avoid any uncompleted transferred file. Those files are stored into a dedicated memory area.
Step 317: a ‘USERS’ database related to all local LogLookers is created from the files previously stored. Below is an illustration of the ‘USERS’ database:
Step 318: the ‘employee’ database is updated. The purpose of this database is to identify the userid of the employee recorded in the violation database. An illustration of the employee database is shown below:
Step 319: the ‘DBLOG’ central database updated on step 314 with the imported information is browsed.
Step 320: a verification is made that the violation userid is linked to an existing entry of the ‘employee’ table.
Step 321: If this userid is not found in the ‘employee’ table, a flag ‘Not found’ is added in the violation record. This information is used to evaluate the level of warning.
Step 322: If the userid is found in the ‘employee’ table, the corresponding e-mail address is added to the violation record. The information may be used to contact the owner of the userid.
Step 323: a verification is made to check if the userid exists in the ‘USERS’ table and defined as an authorized user for the local LogLooker. This information is used to evaluate the level of warning. If the userid is normally defined, a lower weight is attributed to the violation and a ‘low’ violation warning is issued.
Step 324: if the userid doesn't have any right on the violated local LogLooker, a flag ‘Unauthorized’ is added in the ‘DBLOG’ table. The level of warning for this type of violation is considered as a medium one and a ‘medium’ warning is issued.
Step 325: if the userid exists on the local LogLooker violated machine, then the description field of the violation database ‘DBLOG’ is updated with the description of the userid existing into the local LogLooker.
Step 326: a verification is made to check if the userid is present in the suspect list of the ‘SPY’ table. This is one another important feature of the present invention. This table contains all default userids existing in Operating Systems as well as in applications (root, db2admin, administrator, teacher, etc . . . ). This list is continuously updated. Those userids are used by most Worm viruses trying to copy themselves into the Servers.
Step 327: if the userid is found in the ‘SPY’ table, a ‘found’ information is added to the violation record. A high level of warning is attributed in the ‘DBLOG’ violation table.
Step 328: the ‘DBLOG’ table is read for the newest records.
Step 329: each name of an attacker device is compared with a list of devices names that are authorized by the company rules to be in violation. If the name is not found, next step is 332.
Step 330: if the name is found in the list of authorized device names, the administrator is informed of the details related to the found device name and an ‘inform’ level of warning is issued.
Step 331: Those authorized violations are removed from the foregoing analysis processing. And the process continues with step 332.
Step 332: The attacker userid is compared to the ‘employee’ table and to the ‘USER’ table. If not found, next step is 335.
Step 333: If the attacker userid exist in the two mentioned tables, the administrator is informed of the result of step 332 and a ‘low’ level warning is issued.
Step 334: Those low level violations are removed from the foregoing processing.
Step 335: The attacker userid is compared to the suspect ‘SPY’ table. If not found, next step is 338.
Step 336: If the attacker userid is found, the administrator is informed of the result of step 335, and a ‘high’ level of warning is issued. This high level warning contains all suspect bad-logons ordered by userids.
Step 337: Those high level violations are removed from the foregoing processing.
Step 338: The attacker userid is compared to the ‘employee’ table and to the ‘USER’ table. If the userid is found in the ‘employee’ table but not in the ‘USER’ table, next step is 339, otherwise the process goes to step 341.
Step 339: the administrator is informed of the result of step 338 and a ‘medium’ level warning containing all unauthorized bad-logons ordered by userids is issued.
Step 340: Those medium level violations are removed from the foregoing processing.
Step 341: All previous issued warnings (inform, low, medium, high) are collated and sorted to be grouped by attacker device names into final reports. The administrator is thus immediately aware that one attack is happening even though only one or two violations are done on several attacked devices. This correlation between the suspect identified userid and the other violations (even though if normal) clearly shows the suspicion of a virus attack.
Step 342: Those ‘inform, low, medium and high’ level violations are removed from the foregoing processing. The overall process restarts from step 301.
It is to be noted that the advantage of the present invention is to group all the violation information into one relational database for all Operating Systems allowing the previously described cross-checking. It is to be appreciated that many modifications may be made to the method, system, and computer product herein described without departing from the spirit of the invention.
| Number | Date | Country | Kind |
|---|---|---|---|
| 04292461.3 | Oct 2004 | EP | regional |
