Network flow quick security check engine towards content
Technical field
The present invention relates to a kind ofly be particularly related to towards content aware network traffics detection technique specifically a kind of network flow quick security check engine towards content.
Background technology
Network security is seriously to perplex the biggest obstacle of network and information application always; Though industry is continually developed diverse network security tools such as building fire compartment wall, intrusion detection, IPS, anti-virus, anti-Trojan, anti-fishing for a long time; Also obtained effect preferably; But the high chi evil spirit in road is high one zhang, and diverse network harm software constantly updates, constantly makes a variation, constantly develops, and becomes worse greatly.And along with the improving constantly of the network bandwidth, and the continuous expansion of network size and userbase, network traffics are explosive growth just apace, thereby the diverse network security tool is proved definitely inferior on performance, are difficult to deal with.
More serious is, compares with network infringements such as traditional hacker, virus, wooden horses, and the fail safe of content is to human, bigger to the destructive power of society.Such as various yellow information, yellow phonotape and videotape, yellowly play, give currency to stories of sex and violence, network defraud, network gambling, harmful propagate or the like, had a strong impact on social safety, national safety, household safe, individual health of human body.Unfortunately, the instrument of these harmful Web contents of control processing is but very limited, and relevant at present practical product also is in that stage of very elementary development, and is not only fairly simple on function, more very low on performance.
Solve the function and the bottleneck of performance problem of network security tool, except the higher algorithm of the stronger efficient of development function, main approach is to adopt artificial intelligence and software systems hardwareization technology.The present invention is exactly a further investigation on this developing direction and puts into practice; Employing based on the Characteristic Recognition of strategy with traffic classification; Handle through the semantic information of multimedia Characteristic Recognition of pipeline system on this basis, realize degree of depth identification the fast detecting and the content of network traffics.
Summary of the invention
The purpose of this invention is to provide a kind of network flow quick security check engine towards content.
The objective of the invention is to realize by following mode, mainly by network flow characteristic detect, Web content pipeline system depth detection with processing, be harmful to the statistical analysis of content flow person's movements and expression, traffic characteristic and content characteristic storehouse (characteristic control register) four big subsystems and form.
The groundwork step is following:
Network traffics are input to the buffer of " stream/detection of packets subsystem "; Detector carries out matching detection according to the various characteristic segments while convection current/packet package in the feature register and (mainly detects form and type attribute; Do not detect payload content), detected type is divided into secure and trusted, harmful, uncertain three major types unusually.According to the processing policy in the strategy controller, harmful unusually stream/grouping is discarded into dustbin then, directly delivers to green put-through channel safe and reliable; The uncertain content depth that sends to is detected processing subsystem.
(1) content detection processing subsystem at first is stored in uncertain content and detects the input-buffer district, detects identification through text detector, voice detector, visual detector and related detector, is transferred to the information filtering buffer area then.
(2) at the information filtering buffer area; The relevant identification information of sending according to each detector of front by text filter, acoustic filter, picture filter; Implement convection current/grouping and carry out filtration treatment, will be harmful to content and be dropped into dustbin, good content is sent to the flow synthesizer.
(3) the flow synthesizer is responsible for spreading with the information on the green channel with through the green that the good information of filtration treatment is integrated into " secure and trusted " and is delivered in the network of back.
(4) the content characteristic control register is deposited the semantic feature control section (these characteristic segments have a lot of bars) of various harmful contents, supplies each detector/filter network flow to be handled on content detection and filtration streamline as operating basis.
(5) stream characteristic control register is deposited a series of pattern information (characteristic segments) and tactful control information about all kinds of websites, flow, packet package security feature; Deliver to the feature register and the strategy controller of stream/detection of packets subsystem, so that control is to the processing of network traffics.
(6) unusual person's movements and expression statistical analysis subsystem, accept the secure ID information of the various streams that each content detector sends---for example from the Health Category of content in certain website or URL (message address) flow, through statistical analysis after a while; Judge the safety or the health index of all kinds of streams; When surpassing preset threshold, promptly send corresponding report, the report reciever can be back-stage management person; Also each characteristic control register, strategy are controlled device, and reciever is Correction and Control information in view of the above.
Excellent effect of the present invention is following:
(1) flow safety detection and content detection are filtered comprehensively to be the exchange architecture of one, to have improved the fiting effect between the two, reduced many coupling links between the two, can effectively improve processing speed, reduce and realize cost;
(2) stream/packet package detector makes follow-up scale to content detection lower greatly, thereby increases substantially contents processing speed through flow is implemented the classification processing based on strategy;
(3) based on the content detection and the Processing Structure of feed forward type pipelining; Make that the detector operating process of each medium dimension is very fast simple---be particularly suitable for concurrent on a large scale vector calculus or matrix operation; And the result that will detect is ahead of seized packet package and delivers to content filter, can effectively accelerate the processing speed to dividing into groups;
(3) setting of characteristic control register has realized control and has operated in structural independence, and the system that guaranteed has characteristics such as constantly evolution, constantly expansion, constantly upgrading, flexible configuration;
(4) design of unusual person's movements and expression analytic function can make engine have the function of continuous self-study, helps improving constantly effect and the efficient that detects identification.
Description of drawings
Accompanying drawing 1 is the structural representation towards the network flow quick security check engine of content.
Embodiment
Explanation at length below with reference to accompanying drawing the network flow quick security check engine towards content of the present invention being done.
Network flow quick security check engine towards content of the present invention; Based on network traffics detecting processing system of the present invention; Can unify to realize efficiently that network flow monitoring function and bad Web content detect filtering function; And can under relatively low cost situation, improve treatment effeciency exponentially.Particularly along with to the research of network traffic content identification deeply with the more proposition of good algorithm, can realize upgrading expansion continuously, thereby can dependable flow detect the desirability of effect and information filtering effect system.Such system is particularly suitable for big-and-middle-sized communication network, big-and-middle-sized enterprises and institutions network, big-and-middle-sized internet node, catenet safety detection center, large-scale Internet bar; And to the various places of performance requirement height to harmful content erotic, so the present invention is with a wide range of applications and good practical value.
Embodiment
Detect and information filtering engine or system based on network traffics of the present invention; Can adopt Host Based software to realize, also can adopt very lagre scale integrated circuit (VLSIC) (like network flow quick security check engine) to realize with hardware towards the network flow quick security check engine SIC of content, FPG towards content.A kind of implementation in back is main developing direction, and method and points for attention when specifically realizing are following:
(1) each subsystem is preferably on the same chip and realizes, particularly flow detector, content detection and filtration, characteristic control register three sub-systems are placed on the adjacent position of same chip as far as possible, so that guarantee that operation is high-speed;
(2) the characteristic control register is deposited stream/grouping feature control field, and the quantity of these fields may be very big.If the quantity of feature field is big (such as above several ten thousand) too; Can consider characteristic control child register is divided into two parts; A part is placed on the chip at detector/filter place, and another part is placed on the other chip (particularly those characteristic segments that are of little use relatively, the processing time is required undemanding relatively characteristic segments);
(3) unusual person's movements and expression analyzing subsystem also is placed on the same chip as far as possible, because the discrepancy holding wire of this module is many, is put into the outside inconvenience;
(4) dustbin generally will be put into chip exterior.Particularly when system is constantly ripe, very accurate owing to the detection of flow is handled, so dustbin can remove;
(5) the flow synthesizer is fairly simple, on the shared sheet resource seldom, so should be put on the sheet as far as possible;
(6) other part of system such as power supply, network interface, debugging/monitor-interface, management software etc. can be with reference to the network device processing of maturation.