Distributed filesystem network security extension

CA 02525249 2005-11-09

WO 2004/104902

PCT/GB2004/001629

DISTRIBUTED FILESYSTEM NETWORK SECURITY EXTENSION

Field of the Invention

The present invention relates in general to network systems and in particular to distributed filesystems. Still more particularly, the present invention relates to security features for access to distributed filesystems.

Background of the Invention

In general purpose computing systems, such as those supporting versions of the Unix operating system (OS), applications may access data stored on disk drives by means of a set of operating system services including a filesystem (Unix is a registered trademark, exclusively licensed through the Open Group). A filesystem may be employed by a computer system to organise a large collection of files into individual files and directories of files and to map those files to storage devices such as disks. Filesystems comprise two primary components, the programs that control the physical representation of the files and the files themselves that are stored on the disk.

In a distributed computing environment, a number of computing systems can be interconnected by way of a communication network or other coupling facility and can share files by way of a distributed filesystem.

A filesystem exporter is typically executed on the server node (the computing system that controls access to the disk containing the filesystem data), while a filesystem importer is typically executed on the client nodes (other computing systems utilised to access the files on the disk). Accesses to shared files by users on the client nodes are referred to as remote accesses. Accesses to shared files made by users on the server node are referred to as local accesses.

The network filesystem is stored on a server or node of a network, and the server or node is accessible from client terminals (i.e., user computers) that are typically remotely linked to the network. The actual link may be a wired link, as in a standard Ethernet-based local area network (LAN) or a wireless link, such as a Bluetooth Virtual Private Network (VPN). The process of accessing the filesystem via the client terminals is referred to as mounting a filesystem. When a filesystem is mounted, the filesystem's control program reads certain information from the disk concerning the layout of filesystem objects. From this

CA 02525249 2005-11-09

WO 2004/104902

PCT/GB2004/001629 information, the filesystem constructs data structures known as virtual filesystems or Vfs's. Each time a file is opened, or made accessible, the filesystem creates a data structure, referred to as a vnode, which is chained to the vfs.

Each vnode contains information about a given file and contains references to physical file system data structures. The physical file system data structures contain information such as the owner of the file, the size of the file, the date and time of the file's creation and the location of the blocks of the file on the disk. Filesystems include internal data, called meta-data, to manage files. Meta-data may include data that indicates: where each data block of a file is stored; where memory-modified versions of a file are stored; and the permissions and owners of a file.

With more and more companies using remote/network-accessible distributed filesystems to electronically store and later retrieve files/documents, including some with sensitive information, security of distributed filesystems is becoming increasingly important. The IP Security (IPSec) suite of standards was introduced and provides two primary security features: authentication and encryption. In other words, IPSec ensures that sending and receiving machines really are what they claim to be, and IPSec enables data to be scrambled in flight so the data will be incomprehensible if intercepted.

Most systems thus require an authentication of the user during the initial mount, which typically includes verifying user-passwords, etc. However, password-protection and similar security measures are notorious for being open to cracking and can easily be compromised, and the industry has recognised that password-protected systems offer very little protection to sensitive files once general access to the filesystem is obtained.

More advanced hackers also gain access to the files stored on the filesystem by tapping into a transmission during an authorised mount and simply copying the data as it is being transmitted from filesystem to client system. This occurs because, with most password-protected distributed filesystems, once the several levels of security log-in (password verification, etc. are completed, the actual transmission of the files from the filesystem occurs in clear text. Thus, when the transmission includes very sensitive data, additional security measures

CA 02525249 2005-11-09

WO 2004/104902

PCT/GB2004/001629 are required to ensure that the clear text data is not available by simply copying the file during transmission.

The ease at which the security of the sensitive information may be compromised via this latter method depends to some extent on the medium being utilised by authorised users to mount/access the filesystem. For example, wireless access/transmission is typically more prone to eavesdropping and cracking that wire-full (wired) network media. However, even the standard Ethernet can easily be breached without detection, and thus the standard Ethernet is also an unsafe option for routing sensitive data.

As mentioned above, the industry has responded to the growing need for security on the transmission medium by imposing heavy encryption on all transmitted data during a mount of the filesystem. Currently, there are several encryption algorithms and standards (e.g., wireless transport layer security) designed to provide security for the transmissions between client system/node and the server hosting the filesystem. Utilization of heavy encryption requires placing a heavy processing burden on the client system and the server for all traffic. The overall performance of the system is degraded, and significant costs are incurred by companies that wish to implement system-wide encryption for access to their filesystem. Encryption is built into the communication mechanisms and applied to all traffic between client system and server although the majority of traffic may not require that level of security (e.g., non-sensitive information/files).

The utilization of wireless systems to access filesystems is increasing as companies provide remote access to users who may be mobile and wish to connect to the network remotely. Wireless connections are, however, more susceptible to cracking than wired connections. Some wireless users use WTLS, but this security feature is known to be a relatively weak level of security. One solution requires a Virtual Private Network (VPN) data encapsulation/encryption to access sensitive data, even when the majority of clients are accessing the filesystem via token ring. This VPN data encapsulation would further negatively impact the speed of the servers as they encrypt and decrypt all data.

It is also possible to configure VPNs or servers on a VPN to recognise IP addresses or subnets and only require encryption on certain subnets. One problem with this solution is that the administrator of the distributed filesystem server must have knowledge of every wireless node that is not within the network. If a wireless network is set up by an

CA 02525249 2005-11-09

WO 2004/104902

PCT/GB2004/001629 organisation within their department, the server administrator would need to be made aware of the wireless network so that the subnet could be added to the VPN list of IP addresses.

In light of the foregoing, the present invention recognises that it would be desirable to have a method, system and data processing system that dynamically implements enhanced mount security when access to sensitive files on a distributed filesystem is requested. A method and system that would automatically provide a secure mount whenever sensitive file/data are about to be accessed during an ongoing session would be a welcomed improvement. It would be further desirable if the secure mount was completed in a seamless manner so that the authorised user receives access to the sensitive file without experiencing a disconnect and re-mount authentication process, while the sensitive file is shielded from unauthorised capture by routing the sensitive file via the more secure mount.

DISCLOSURE OF THE INVENTION

Disclosed is a method, system and computer program product that dynamically implements enhanced mount security of a filesystem when access to sensitive files on a networked filesystem is requested. The client system initiates a standard mount and authentication process for access to files of the filesystem. When the user of the client system attempts to access a specially tagged sensitive file, the server executes a software code that terminates the current mount. The server is re-configured to route to a secure port any attempts to re-mount the server from the IP address associated with the client. When a session is terminated by the server, the client system is programmed to automatically attempt to re-mount the server. The server recognizes the IP address of the client during the remount operation and routes the client to the secure port.

A secure mount is thus automatically provided whenever sensitive files/data are about to be accessed during an ongoing session that was initiated on a standard mount. Then routing via a secure mount is completed in a seamless manner so that the authorized user receives access to the sensitive file without experiencing significant delay or a visible disconnect that requires user-initiated remount and authentication processes. Meanwhile the sensitive file is shielded from unauthorized capture by routing the sensitive file via the more secure mount established.

CA 02525249 2005-11-09

WO 2004/104902 PCT/GB2004/001629

BRIEF DESCRIPTION OF THE DRAWINGS

The invention will now be described, by way of example only, with reference to the accompanying drawings, in which:

Figure 1A is a block diagram of a data processing system within which the features of the invention may be implemented;

Figure IB is a block diagram representation of the files within the filesystem of Figure 1A having a security tag indicating a required level of security according to one embodiment of the invention;

Figure 2 is a block diagram of a distributed network within which features of the invention may be implemented according to one embodiment of the invention;

Figure 3A is a flow chart of the process by which a client is provided access to sensitive files during access via a standard mount according to one embodiment of the invention;

Figures 3B and 3C are flow charts of the processes by which a server monitors and controls client requests for access to sensitive files to ensure that access to those files is routed via a secure channel according to one embodiment of the invention; and

Figure 4 is a block diagram illustrating logic components for seamlessly completing a switch of a client session on a filesystem from a standard, non-secured channel to a secured channel during a single continuing session in accordance with one embodiment of the invention.

DETAILED DESCRIPTION OF THE INVENTION

With reference now to the figures and in particular with reference to Figure 1A, there is illustrated a block diagram of a computer system, which may be utilised as either a server hosting the distributed filesystem or a client system utilised to mount the server at which the distributed filesystem is hosted. Computer system 100 comprises processor 102 and memory 104 connected via system bus/interconnect 110. Computer system 100 also comprises Input/Output (I/O) Channel Controller (CC) 109, which is coupled to interconnect 110. I/O CC 109 provides connection to I/O devices 106, including Redundant Array of Disk (RAID) 114. RAID 114

CA 02525249 2005-11-09

WO 2004/104902

PCT/GB2004/001629 stores instructions and data that are loaded to memory as needed by the applications being executed by the processor. According to the illustrative embodiment, RAID 114 provides the storage media for the plurality of files that constitute filesystem 112.

Computer system 100 further comprises network connection devices 108, which may include wireline modem, wireless modem, and Ethernet card, among others. Access to and from I/O devices 106 and network connection devices 108 are routed through I/O channel controller (I/OCC) 109, which includes logic for completing the automatic re-establishment of a mount to/from computer system 100 via a secure path/channel/port when required, as further described below.

Computer system 100 includes operating system (OS) 122, filesystem software application 124, mount code 125 and Distributed Filesystem Network Security Extension (DFNSE) 126. Filesystem software application 124 provides the basic accessing, maintaining, and updating of filesystem 112, when computer system 100 is being utilised to host a filesystem 112.

When within a client system, filesystem software application 124 includes client version of mount code 125 for completing a mount and automatic re-mount of the server hosting the filesystem. In the illustrative embodiment, the automatic remount process is implemented by the client system whenever an established mount with the server is disrupted/lost without the client having completed an unmount of the server. In the described embodiment, the server may issue a FYN command to terminate a current mount and thus force the client to initiate a re-mount of the server. The FYN command is issued in response to access to particular files that require special security protections, as will be explained in greater detail below.

Returning to Figure 1A, and the description of filesystem software application 124, when executed within a server, filesystem software application 124 includes code for receiving, maintaining, and verifying credential information of various users and client systems, code for maintaining filesystem 112 and code for selectively initiating a security software and associated response, called Distributed Filesystem Network Security Extension (DFNSE) 126. DFNSE 126 provides the backbone of the inventive features herein and the execution of DFNSE 126 on a server is described below with reference to Figures 3A-3C and 4A. At a basic level, DFNSE 126 determines what level of security access is permitted/authorised

CA 02525249 2005-11-09

WO 2004/104902 PCT/GB2004/001629 for specific files of the filesystem 112 and when to initiate the enhanced security measures of the invention.

With reference now to Figure 2, there is illustrated an example network comprising multiple interconnected computer systems, which may be similarly configured to computer systems 100 of Figure 1 that are advantageously utilised to provide either the server or client functionality for respectively hosting or accessing a distributed filesystem. Network 200 comprises a distributed filesystem 202 hosted on three (or more) interconnected servers 203. Network 200 also comprises a plurality of client systems 201 connected to distributed filesystem 202 via a network backbone 210. Network backbone 210 comprises one or more network-connectivity systems (or sub-networks) that may be configured according to network protocols such as Ethernet or TokenRing. These sub-networks may be, for example, wire-line or wireless local area networks (LAN) or a wide area network (WAN), such as the Internet. Additionally, sub-networks may include fibre optic network as well.

Distributed filesystem 202 is directly coupled to network backbone 210 via one or more ports (not shown) on at least one of the servers 203. Client systems 201 may be either directly coupled to the network backbone (wireline) or communicatively connected via a wireless medium illustrated by wireless antenna 207. Client systems 201 access the distributed filesystem 202 via one of the various available media utilising one of the various network configurations, each one having a different level of susceptibility to cracking. Thus, client system 201 may access and mount filesystem 202 via a non-secure wireless network 227, or client system 201 may mount filesystem 202 utilising a secure fibre-optic network 225. For simplicity of describing the invention, the wireless network 227 will be assume to be a standard, non-secure network without encryption, while the fibre-optic network 225 is assumed to be a special, secured connection with encryption. Each connection is routed via a different one of the ports available to the server 203 at which the mount of filesystem 202 is supported.

Figure IB illustrates a block diagram representation of filesystem 202 with a more detailed delineation of the files that comprise filesystem 202. As illustrated, filesystem 202 comprises a control block 131 and a plurality of files 132a-n, each of which includes a metadata tag 112 with header/identifier field 334 and security field 336. · Header/ID field 334 contains information about the file ID and the users who have access to

CA 02525249 2005-11-09

WO 2004/104902

PCT/GB2004/001629 the file. Security field is a single bit field, which indicates the level of security attributable to that file and consequently the type of user-access permitted. According to the illustrative embodiment, certain files that require highest levels of security and which are restricted to being accessed solely on a secured mount (e.g., files 1 and 3) are tagged with 1 in the security field 336 of their respective metadata. Other files not so tagged (i.e., tagged with a 0) are normal (e.g., file 2) and may be access by any authorised user without a special secured mount.

As mentioned above, the invention introduces an enhanced security mechanism, which, in the illustrative embodiment, is referred to as DFNSE (Distributed Filesystem Network Security Extension). With DFNSE, the filesystem server is able to infer from the file permissions associated with a file or directory the level of network security that is required when providing access to the file by particular users. DFNSE is a server-level filesystem security enforcement application and/or procedure. Accordingly, with DFNSE, only the server is required to have knowledge of the networks connections or adapters being utilised by the server.

Specific hardware, logic and software components are provided within each server capable of providing a mount to the filesystem to implement DFNSE. Figure 4 is a block diagram illustrating some of these components. As illustrated in Figure 4, server 203 may be provided with two Ethernet network adapters (or ports): enO 403, which is secure, and enl 405, which is not secure. In one embodiment, the network topology behind these adapters is consistent. That is, the sub-network selected itself provides the security and the server is able to dynamically select between sub-networks based on the level of security required. In another embodiment, additional encryption or other security features are provided with the secure adapter, enO 403.

EnO 403 connects to a fibre network 225 that is utilised to route all sensitive data, while enl 405 connects to a standard Ethernet-based wired network 221 and is utilised for routing all other (non-sensitive) data communication. Enl 405 is the default port for mounting the filesystem's server. The exemplary embodiment assumes that the ease at which the standard Ethernet can be breached without detection makes the Ethernet an unsafe option for routing sensitive data. During a mount of the filesystem, the server, which has detail knowledge of the file permissions 334 and security level 336 for each file tracks user access and determines when to force the client to switch over to the secure network based on the file permissions in place for the files being

CA 02525249 2005-11-09

WO 2004/104902

PCT/GB2004/001629 accessed. According to the illustrative embodiment, the network topology is consistent for both the secure and non-secure routes so that no additional hardware and/or routing protocol upgrades are required to account for different topologies during the switch from non-secured to secured sub-network.

Server 203 of Figure 4 also includes a mount controller 407, which performs conventional mount support and unmount operations for filesystem 202 as well as re-mount configuration in accordance with the features of the present invention. Mount controller 407 includes DFNSE 126 and is preferably embodied as software code executing on server 203. DFNSE 126 operates to trigger mount controller 407 to route a request for a mount via either standard port, Enl 405, or secured port, EnO 403. Server 203 also includes encryption module 409 utilised in conjunction with DENSE 126 and EnO 403, when encryption is implemented on secured port.

During filesystem access, when a request for access to sensitive operation is received at server 203, mount controller 407 marks the client's IP address as one needing access to sensitive data. Server 203 then breaks the current connection, (i.e., the server sends a FYN to the client). The client automatically attempts to reconnect, and mount controller 407 recognizes the client IP address during the re-mount. The client's session is then directed to a secure SSL port. Thus, while primary access is provided via the standard port, access is dynamically switched to the SSL secure port when access to sensitive data/files is required.

Turning now to Figure 3B, there is illustrated a flow chart of the process by which the software-implemented DFNSE security features are implemented within the above hardware/logic configuration of the server hosting the filesystem. The process begins with a standard mount request received at the standard port of the server from a client as shown at block 321. The user is prompted for authentication data (password, etc.) and the client system's IP address is retrieved from the data packet as indicated at block 323. A session is opened on the standard port and both the IP address and the user's authentication data are stored within a parameter file linked to the particular session as shown at block 325.

Once the session is established, i.e., server logic associated with access permissions, etc., monitors the user's interaction as shown at block 327 and a determination made at block 329 whether the user is requesting access to a sensitive file.

CA 02525249 2005-11-09

WO 2004/104902

PCT/GB2004/001629

The server that is satisfying a clients request for a file is programmed with the authorization/credentials of the user on the remote client and the permissions of the file being accessed. If the file being accessed is not sensitive, regular access is provided to the user on the standard port as shown at block 331. However, when the file being requested by the client is a sensitive file that requires a more secure channel before access can be granted, a next determination is made at block 333 whether the user has proper access permission to access the file. If the user does not have proper access permission the request is denied as shown at block 335. If, however, the user's credentials indicates the user has permission to access the particular file, the DFNSE security protocol is activated as shown at block 337. Activation of DFNSE causes the server to force an unmount of the client by issuing a FYN to the client and concurrently configuring a more secure port to accept the re-mount from the client having the IP address saved with the session parameters. The server then provides secured access to the file via the secured port as shown at block 339.

Figure 3A is a flow chart of the processes involved in the implementation of the invention primarily from the perspective of the user/client system. The process begins when a user (via the client system) first mounts a server hosting the filesystem as illustrated at block 302 and requests access to the filesystem as shown at block 304.

When the client initially mounts the NFS filesystem, the mount is completed over a standard TCP connection by default. For example, the connection may be to the well known NFS port of 2048. The server has a listening socket bound to this port and operates according to the standard (non-secure) protocol. Notably, in the illustrative embodiment, the standard protocol is enhanced by the DFNSE protocol, which is implemented when access to sensitive files are requested.

When a connection is requested from the client, the listening socket of the server basically duplicates itself and bounds the connection to the remote client. The listening socket then remains open to handle other connection requests. Authentication of the client is initiated as shown at block 306 and a determination made at block 308 whether the client's authentication was successful. If the client/user authentication process is unsuccessful, access to the filesystem is denied and the mount is disconnected as shown at block 310. Then the process ends as indicated at block 311. Otherwise, a session is opened and the user is provided access to the filesystem as shown at block 309.

CA 02525249 2005-11-09

WO 2004/104902

PCT/GB2004/001629

The client system monitors the connection for disconnects as shown at block 312 and determines as indicated at block 314 whether the connection becomes unresponsive or is prematurely broken at the server side (i.e., ideally when a server issued FYN is received). When the connection becomes unresponsive or broken, the client initiates a remount that is routed to the port indicated by the server as shown at block 316.

Notably, reconnection in response to a server-initiated dismount is directed on a secure port at the server, although the actual port may be unknown to the client system. Utilizing the security protocols of DFNSE and based on the knowledge of which port is secure and whether the session requires a secure port, the server is able to request that the client re-mount over a secure port. For example, the client may be made to re-mount utilizing a port running Secure Socket Layer. Notably, no user-action is required to complete the re-mount and port-switching procedures. The monitoring for server-side unmount and subsequent re-mount all occur as background processes at the client system, and the user (client) is not made aware of the switch to a more secure port.

A more detailed account of the internal processing required for rerouting via a more secure port at the server side is illustrated by the flow chart of Figure 3C. The process begins when access to a sensitive file is identified by DFNSE as shown at block 351. The server checks the current port security as indicated at block 352. A determination is made at block 354 whether the current port security is sufficient for accessing the requested file (depending on how sensitive the file is, which is deduced by reading the security bit of the file in the illustrative embodiment). If the current port security is sufficient for accessing the requested file, the access is provided as shown at block 356.

In one alternate embodiment, the remount function may be selectively automated and the process would require a next determination whether the feature for automatic remount is enabled. With this alternate embodiment, if the automatic remount capability is not enabled, the user will actually be prompted to remount via a secure mount.

Returning to the illustrated embodiment of Figure 3C, when the port security is insufficient, the server responds by selecting a more secure port (e.g., EnO) for the session as shown at block 358. The server takes a snapshot of the authentication and mount parameters of the session, including the client's IP address, and transfers these parameters to the control logic of the more secure port as depicted at block 360. The transfer occurs with very little latency and the more secure port is thus

CA 02525249 2005-11-09

WO 2004/104902 PCT/GB2004/001629 automatically configured to receive a re-mount from that client and continue supporting the session in progress. After the secure port has been configured, the corresponding port number is given to the mount controller along with the IP address of the client. The server terminates the mount on the first standard port and re-establishes the session via the more secure port when a re-mount is received from the client as shown at block 362.

Notably, in response to the server terminating the initial mount, the client initiates the re-mount which is directed by the user to the second, secured link. This re-establishes the initial session of the client but via the second port. Re-establishing the connection involves checking the clients IP address and matching it to the port that is set up to receive the connection from that IP address. The entire process occurs in the background and thus a seamless switching of ports is completed from the user's perspective.

In one alternate embodiment, the level of security attributable to a particular file is determined by the users (or selected client systems) that are provided access to the particular file. Thus, if file access permissions are restricted to filesystem administrators only, then the security level is high, while file access permission given to regular employees indicates a relatively low level of security required. Determination of the security level for a file is completed when the user initially creates the file and assigns the access permission to that file. Once the file is placed within the filesystem, the file automatically inherits the network security protection that is in place. With this implementation, existing file permissions on files within the filesystem (e.g., UNIX -rwx,rwx,rwx for user, group, other) fold into the security model provided herein without requiring extensive system administration and configuration. Thus, the present invention eliminates the need for reconfiguring existing filesystems on a per file basis. With the invention, there is also no requirement to move sensitive files to a secure server.