A signaling conversion method and system based on telephone interaction

Abstract

The application discloses a signaling conversion method and system based on telephone interaction, which is applied between a service domain and an operator access domain, and the method comprises the following steps: on the service domain side, receiving original signaling messages from an internal service system, performing signaling desensitization mapping processing on the original signaling messages, generating abstract metadata containing only service logic identification, and transmitting the abstract metadata to the access domain side across the domain through a kernel-level communication interface; on the access domain side, converting the abstract metadata into an identifiable signaling format of an operator, and sending the abstract metadata to an operator network; establishing a media stream transmission tunnel at an operating system kernel layer, and monitoring transmission quality indexes of the media stream transmission tunnel in real time; continuously monitoring a signaling registration transaction log on the service domain side through an independent relay state maintenance module running independently of the internal service system, caching complete parameters of the latest successful registration to the operator network based on the log, and realizing high-safety, high-reliability and high-compatibility signaling conversion.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and more specifically to a signaling conversion method and system based on telephone interaction. Background Technology

[0002] With the increasing prevalence of cloud computing, more and more internal business systems (such as bank IVR and call centers) are being deployed in public or hybrid cloud environments. These systems need to interact with the traditional IMS network of operators to enable telephone call functionality.

[0003] Currently, traditional solutions typically use SIP (Session Initiation Protocol) gateways for protocol conversion. However, traditional gateways usually parse and forward SIP signaling from both internal and external networks simultaneously in a single processing unit or process. This allows malicious signaling packets crafted by attackers to bypass simple filtering rules and directly exploit protocol stack vulnerabilities to attack internal business systems, resulting in insufficient security isolation. Furthermore, in a NAT (Network Address Translation) environment, NAT drift can cause changes in the media stream's four-tuple (source IP, destination IP, source port, destination port). Traditional solutions require triggering SIP renegotiation (Re-INVITE), causing a call interruption of several seconds and resulting in insufficient media stream stability. Additionally, since internal business systems maintain connections with the IMS core network through a unified heartbeat, if an internal business system crashes or experiences jitter, heartbeat timeouts will cause the operator to reset the trunk link, with recovery times reaching several minutes, resulting in low trunk link reliability. Moreover, due to differences in SIP protocol implementations (such as header fields and interaction frequencies) across different operator IMS networks, internal business systems need to be custom-developed for each operator, leading to high development and maintenance costs and complex multi-operator adaptation. Summary of the Invention

[0004] To overcome the shortcomings of the prior art, the present invention provides a signaling conversion method and system based on telephone interaction to solve the problems in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A signaling conversion method based on telephone interaction, applied between the service domain and the operator access domain, includes:

[0007] On the business domain side, the system receives raw signaling messages from internal business systems, performs signaling desensitization mapping processing on the raw signaling messages, removes network topology sensitive fields, generates abstract metadata containing only business logic identifiers, and transmits the abstract metadata across domains to the access domain side through a kernel-level communication interface.

[0008] On the access domain side, the abstract metadata is converted into a carrier-recognizable signaling format and sent to the carrier network;

[0009] A media stream transmission tunnel is established at the operating system kernel layer. The transmission quality indicators of the media stream transmission tunnel are monitored in real time. When the transmission quality indicators meet the preset degradation conditions, a dynamic drift self-healing process is triggered. The dynamic drift self-healing process includes redirecting the media stream to the optimal path calculated based on the quadrupole correlation matrix and path scoring algorithm through the eBPF program, so as to complete the switching without interrupting the existing media session.

[0010] An independent relay status maintenance module, operating independently of the internal business system, continuously monitors the signaling registration transaction logs on the business domain side. Based on the log cache, it caches the complete parameters of the most recent successful registration with the operator network. When the internal business system fails, it independently maintains the registration status and heartbeat with the operator network using the cached registration parameters.

[0011] In one embodiment, signaling desensitization mapping processing is performed on the original signaling message, including:

[0012] Sensitive fields in the original signaling message are identified using a preset sensitive field identification rule base;

[0013] The abstract metadata is generated by replacing the specific values ​​of the sensitive fields with a preset mapping template.

[0014] The sensitive fields include at least the IP address in the Via header, the terminal identifier in the Contact header, and the user identity information in the From / To header.

[0015] In one embodiment, the kernel-level communication interface is a socket based on the Linux kernel AF_NETLINK or AF_XDP, which is bound to the signaling network card and media network card of the dual network card server, and the signaling and media data are classified and isolated for transmission through the BPF program filter.

[0016] In one embodiment, the transmission quality metrics include packet loss rate, latency, and transmission quadruple, and the preset degradation condition includes at least one of the following:

[0017] The average packet loss rate is greater than 10% over three consecutive sampling periods.

[0018] The media stream has a one-way latency exceeding 200ms and lasts for more than 500ms.

[0019] A sudden change occurs in any of the four elements of the media stream: source IP address, destination IP address, source port, and destination port.

[0020] In one embodiment, redirecting the media stream to the optimal path calculated based on a quadruple association matrix and path scoring algorithm via an eBPF procedure includes:

[0021] The mapping relationship between media stream quadruples and transmission paths is maintained in real time through the quadruple association matrix manager;

[0022] When transmission quality degradation is detected, the optimal path is calculated based on the current network topology and load status using a path scoring algorithm. The scoring factors of the path scoring algorithm include real-time network metrics, historical stability, operator priority, and cost weight. The real-time network metrics include the real-time packet loss rate, latency, and jitter of the path.

[0023] The media stream is redirected to the optimal path via an eBPF procedure.

[0024] In one embodiment, the quadruple association matrix manager includes:

[0025] The main hash table stores the currently active quadruple-path mappings and uses the RCU mechanism for lock-free updates.

[0026] Alternate hash table: used to cache historical four-tuple-path mappings, providing a fallback option in case of abnormal path switching;

[0027] The update triggering conditions for the primary hash table and the backup hash table include at least one of NAT mapping changes, routing table updates, and manual switching instructions.

[0028] In one embodiment, the signaling registration transaction log on the business domain side is continuously monitored, including capturing successful transactions of REGISTER requests in real time through at least one of the system log interface and shared memory ring buffer.

[0029] In one embodiment, it further includes:

[0030] When a fault is detected in the internal business system, corresponding operations are performed based on the fault severity level; wherein, the fault levels include Level 1 fault, Level 2 fault, and Level 3 fault, where Level 1 fault is less severe than Level 2 fault, and Level 2 fault is less severe than Level 3 fault. The corresponding operations performed based on the fault severity level include:

[0031] When the severity level of the fault is Level 1, a temporary response indicating that the call has been queued is returned to the user terminal, and the call is queued to wait for the internal business system to be restored.

[0032] If the fault severity level is level two, the call will be redirected to the emergency IVR module;

[0033] When the fault severity level is three, a pre-recorded fault prompt tone is played to the user, and the relay link registration status is maintained.

[0034] In one embodiment, the step of converting the signaling message on the access domain side further includes a protocol compliance completion sub-step, which includes:

[0035] Based on the target operator's identifier, dynamically load the corresponding protocol completion plugin from the protocol plugin library;

[0036] The protocol completion plugin is used to perform compliance checks and corrections on the converted standard signaling messages; wherein the correction operations include at least one of adding missing SIP header fields, adjusting interaction frequency parameters, and correcting encoding formats.

[0037] A signaling conversion system based on telephone interaction, applied between the service domain and the operator access domain, includes:

[0038] An asymmetric heterogeneous protocol stack architecture, including a business domain protocol stack and an access domain protocol stack;

[0039] The business domain protocol stack is deployed on the business domain side and is used to receive raw signaling messages from internal business systems;

[0040] The signaling desensitization mapping module is used to perform signaling desensitization mapping processing on the original signaling message, remove the network topology sensitive fields, generate abstract metadata containing only business logic identifiers, and transmit the abstract metadata across domains to the access domain side through the kernel-level communication interface.

[0041] The access domain protocol stack is deployed on the access domain side and is used to convert the abstract metadata into a signaling format that can be identified by the operator and send it to the operator's network.

[0042] The media tunnel management module is used to establish a media stream transmission tunnel at the operating system kernel layer, monitor the transmission quality indicators of the media stream transmission tunnel in real time, and trigger a dynamic drift self-healing process when the transmission quality indicators meet the preset degradation conditions. Without interrupting the existing media session, the media stream is switched to the optimized path. The transmission quality indicators include packet loss rate, latency and transmission quadruple.

[0043] An independent relay status maintenance module operates independently of the internal business system. It continuously monitors the signaling registration transaction logs on the business domain side and caches the complete parameters of the most recent successful registration with the operator network based on the logs. When the internal business system fails, it independently maintains the registration status and heartbeat with the operator network using the cached registration parameters to ensure the relay link remains available.

[0044] Compared with existing technologies, the beneficial effects of this invention are as follows:

[0045] By adopting an asymmetric heterogeneous protocol stack architecture, signaling conversion and isolation, dynamic media stream optimization, and independent maintenance of trunk status are integrated to achieve highly secure, highly reliable, and highly compatible signaling conversion. This solves many problems associated with traditional signaling conversion, such as protocol attack risks, media stream interruption, single point of failure caused by service and trunk coupling, and complex multi-operator adaptation.

[0046] Compared to traditional SIP gateway solutions, this invention effectively reduces the attack surface through an asymmetric heterogeneous protocol stack architecture, reduces the media stream interruption rate to below 0.1% (switching latency ≤50ms), ensures the relay link remains available when internal business systems fail, and effectively shortens the development cycle for multi-carrier adaptation. Attached Figure Description

[0047] Figure 1 A flowchart illustrating a signaling conversion method based on telephone interaction, provided as one embodiment of the present invention;

[0048] Figure 2 A schematic diagram illustrating the working principle of a signaling conversion system based on telephone interaction, provided as one embodiment of the present invention;

[0049] Figure 3 This is a technical flowchart of a signaling conversion method based on telephone interaction, provided as one embodiment of the present invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] like Figures 1 to 3 As shown, the present invention provides a signaling conversion method based on telephone interaction, applied between the service domain and the operator access domain, comprising:

[0052] S100. On the business domain side, receive the original signaling messages from the internal business system, perform signaling desensitization mapping processing on the original signaling messages, remove the network topology sensitive fields, generate abstract metadata containing only business logic identifiers, and transmit the abstract metadata across domains to the access domain side through the kernel-level communication interface.

[0053] In one embodiment, signaling desensitization mapping processing is performed on the original signaling message, including:

[0054] S110. Identify sensitive fields in the original signaling message through a preset sensitive field identification rule base;

[0055] S120. Replace the specific values ​​of sensitive fields with a preset mapping template to generate abstract metadata;

[0056] Sensitive fields include at least the IP address in the Via header, the terminal identifier in the Contact header, and the user identity information in the From / To header.

[0057] In this embodiment, addressing the high risk of protocol attacks, traditional SIP messages contain internal network IP addresses and terminal identifiers in header fields such as Via and Contact, forming a potential attack surface. This invention performs signaling desensitization mapping, identifying and stripping (or template-replacing) these sensitive fields to generate abstract metadata that does not contain real network topology and user identity information. This effectively hides information within the business domain, eliminating the risk of sensitive information exposure at the signaling content level. This effectively blocks attackers from using signaling messages for topology probing or targeted attacks, significantly improving system security.

[0058] In one embodiment, the kernel-level communication interface is a socket based on the Linux kernel AF_NETLINK or AF_XDP, which is bound to the signaling network card and media network card of the dual network card server, and the signaling and media data are classified and isolated for transmission through the BPF program filter.

[0059] In this embodiment, the present invention utilizes kernel-level sockets such as AF_XDP, combined with BPF program filters, to achieve rapid processing and routing of data packets at the operating system kernel layer. By binding this interface to independent signaling and media network cards and enforcing data classification and isolation through BPF rules, high performance and low latency in cross-domain transmission of abstract metadata can be ensured. Simultaneously, physical isolation of signaling and media streams prevents mutual interference and cross-domain eavesdropping, further strengthening the system's security boundary and effectively addressing the need for secure and efficient communication between the service domain and the access domain.

[0060] S200: On the access domain side, the abstract metadata is converted into a carrier-recognizable signaling format and sent to the carrier network;

[0061] It should be noted that this invention receives raw SIP messages (such as REGISTER requests) from internal business systems through the business domain protocol stack. Sensitive fields in the raw SIP messages are identified through the signaling desensitization mapping module, and abstract metadata is generated based on a preset mapping template. For example, Contact: sip:user@10.0.0.1:5060 is converted to Contact:sip:abstract_user@bank.com. The abstract metadata is transmitted from the business domain side to the access domain side through the AF_XDP kernel-level communication interface. This interface ensures, through a BPF program, that only abstract metadata can pass through, and the raw signaling is completely isolated. The abstract metadata (i.e., the signaling data without sensitive information carriers) is received through the access domain protocol stack, and the corresponding protocol completion plugin is loaded according to the target operator's identifier. The abstract metadata is completed for compliance based on the protocol completion plugin, such as adding the missing Supported: 100rel header field and adjusting the Session-Expires value. By sending the completed standard SIP message to the operator's IMS network, signaling conversion and isolation are achieved.

[0062] This invention completely blocks protocol-level attack paths by constructing an asymmetric heterogeneous protocol stack architecture and physically isolating and desensitizing signaling between the service domain protocol stack and the access domain protocol stack.

[0063] In one embodiment, the step of converting signaling messages on the access domain side further includes a protocol compliance completion sub-step S210, which includes:

[0064] S211. Based on the identifier of the target operator, dynamically load the corresponding protocol completion plugin from the protocol plugin library;

[0065] S212. Use the protocol completion plugin to perform compliance checks and corrections on the converted standard signaling messages; wherein, the correction operations include at least one of adding missing SIP header fields, adjusting interaction frequency parameters, and correcting encoding formats.

[0066] In this embodiment, the present invention employs a plug-in design, enabling the dynamic loading of specific protocol completion plugins based on the target operator's identifier. These plugins automatically perform compliance checks and corrections on basic signaling messages, such as completing specific header fields and adjusting timer parameters. This eliminates the need for customized core logic for each operator, achieving seamless compatibility across heterogeneous networks and significantly reducing the development and maintenance costs of multi-operator adaptation. It addresses the differences in SIP protocol implementations across different operator IMS networks, thereby resolving the complexities of multi-operator adaptation.

[0067] Furthermore, plugins are loaded via dynamic link libraries, enabling a unified interface, hot reloading, and rollback upon failure, thus improving the flexibility and reliability of the plugin system and facilitating its expansion and maintenance.

[0068] Specifically, after receiving the abstract metadata, the access domain protocol stack parses the target operator's identifier (usually extracted from the Request-URI or Via header of the SIP message). The plugin loader queries the corresponding plugin file (e.g., ChinaMobile_SIP_Plugin.so) from the protocol plugin registry based on the operator identifier. The plugin SO file is dynamically loaded using the Linux dlopen() function. The plugin must implement the unified ProtocolAdapter interface, which includes the validate() and complement() methods. The plugin's validate() method is called to check if the abstract metadata conforms to the operator's basic specifications. The plugin's complement() method is called to complete the signaling message. Common completion operations include adding missing header fields (e.g., P-Asserted-Identity, P-Charging-Vector), adjusting interaction frequency parameters (e.g., changing Session-Expires from 1800 to 3600), and correcting the encoding format (e.g., changing G.711 to G.729). The protocol compatibility testing unit simulates the completed signaling message to ensure it can pass the target operator's IMS device verification. If the plugin fails to load or the verification fails after completion, the system will automatically revert to the default protocol template to ensure basic communication functionality.

[0069] S300: Establish a media stream transmission tunnel at the operating system kernel layer, monitor the transmission quality indicators of the media stream transmission tunnel in real time, and trigger a dynamic drift self-healing process when the transmission quality indicators meet the preset degradation conditions. The dynamic drift self-healing process includes redirecting the media stream to the optimal path calculated based on the quadrupole correlation matrix and path scoring algorithm through the eBPF program, so as to switch the media stream to the optimal path without interrupting the existing media session.

[0070] This step effectively addresses the issue of insufficient media stream stability, particularly call interruptions caused by NAT drift. The transmission quality metrics include packet loss rate, latency, and transmission quadruple. Preset degradation conditions include at least one of the following:

[0071] The average packet loss rate is greater than 10% over three consecutive sampling periods.

[0072] The media stream has a one-way latency exceeding 200ms and lasts for more than 500ms.

[0073] A sudden change occurs in any of the four elements of the media stream: source IP address, destination IP address, source port, and destination port.

[0074] In this embodiment, by setting the aforementioned quantified degradation conditions, the system can accurately and automatically monitor network quality. These thresholds are set for voice communication quality requirements and can promptly identify packet loss, latency, and connection abrupt changes caused by NAT rebinding that lead to a decline in user experience. This provides an accurate triggering basis for subsequent self-healing actions, avoiding the problems of slow response or misjudgment in existing technologies.

[0075] In one embodiment, redirecting the media stream to the optimal path calculated based on a quadruple association matrix and path scoring algorithm via an eBPF procedure includes:

[0076] S310. The mapping relationship between media stream quadruples and transmission paths is maintained in real time through the quadruple association matrix manager; wherein, the quadruple association matrix manager includes:

[0077] The main hash table stores the currently active quadruple-path mappings and uses the RCU mechanism for lock-free updates.

[0078] Backup hash table: used to cache historical four-tuple-path mappings and provide a fallback option in case of abnormal path switching; the update trigger conditions for the primary hash table and backup hash table include at least one of the following: NAT mapping change, routing table update and manual switching instruction.

[0079] In this embodiment, a dual hash table structure (primary table and backup table) combined with the RCU lock-free update mechanism is adopted, enabling the mapping relationship between quadruples and paths to be queried and updated extremely quickly. This ensures that when network degradation is detected, the system can instantly obtain the current stream status and prepare a backup path, laying the data foundation for fast and smooth switching between primary and backup paths. This is a key component for achieving highly reliable media transmission, thereby enabling efficient management of the media stream status.

[0080] Specifically, the quadruple association matrix manager uses a double hash table structure:

[0081] Main hash table:

[0082] Key: A quadruple structure containing source IP (32 bits), destination IP (32 bits), source port (16 bits), and destination port (16 bits).

[0083] Value: The path structure contains the route ID (32 bits), the valid timestamp (64 bits), and the current quality score (floating point).

[0084] Update mechanism: Lock-free updates are implemented using the Linux kernel's RCU (Read-Copy-Update) mechanism. The update frequency is triggered when the network status changes (such as within 100ms after a heartbeat packet is lost).

[0085] Backup hash table: Stores historical quadruple-path mappings, with a capacity of 150% of the main hash table. When path switching is abnormal or the quality of the new path is not up to standard, it can quickly fall back to the best historical path. It adopts an LRU eviction policy to automatically clean up the least used mapping.

[0086] Path prediction submodule: Based on historical path switching records and real-time network topology (obtained via BGP-LS or SDN controller), it uses a Markov chain model to predict sessions that may experience NAT drift within the next 5 minutes, pre-establishes backup tunnels for high-risk sessions, and achieves predictive drift self-healing.

[0087] S320. When transmission quality degradation is detected, the optimal path is calculated based on the current network topology and load status using a path scoring algorithm. The scoring factors of the path scoring algorithm include real-time network metrics, historical stability, operator priority, and cost weight. Real-time network metrics include the real-time packet loss rate, latency, and jitter of the path. The optimal path is calculated based on the mapping relationship.

[0088] S330, Redirect the media stream to the optimal path using the eBPF program.

[0089] In this embodiment, when degradation conditions are triggered, the path scoring algorithm intelligently selects the optimal alternative path by comprehensively considering real-time performance, historical stability, and cost strategies. Most importantly, the media stream is redirected to the new path by directly modifying the packet forwarding rules in kernel space via the eBPF program. Since the entire process bypasses user space and the signaling protocol stack, there is no need to trigger time-consuming SIP renegotiation (Re-INVITE). This allows path switching to be completed in an extremely short time (e.g., <50ms), almost imperceptibly to the user, thereby reducing the call interruption rate caused by network fluctuations to an extremely low level. This effectively solves the problem of long media stream interruption times in existing technologies and achieves lossless media stream migration.

[0090] It should be noted that this invention collects media stream transmission quality indicators in real time, including packet loss rate, latency, jitter, and quadruple status. Transmission quality is determined to be degraded when any of the following conditions are detected:

[0091] The average packet loss rate is greater than 10% over three consecutive sampling periods (100ms per period);

[0092] The media stream has a one-way latency exceeding 200ms and lasts for more than 500ms.

[0093] A mutation occurs in any of the four elements of the quadruple (source IP, destination IP, source port, destination port).

[0094] At this point, the dynamic drift self-healing process is triggered. The quadruple association matrix manager queries the main hash table to obtain the path ID corresponding to the current media stream quadruple. The dynamic drift self-healing algorithm unit calculates the optimal path based on the path scoring algorithm. The scoring factors include:

[0095] Real-time network metrics (packet loss rate, latency, jitter), such as a weight of 40%;

[0096] Historical stability score (based on the number of path changes in the past 10 minutes), weighted at 30%;

[0097] Carrier priority and cost weight, such as a weight of 30%. For example: Total path score = Real-time network metric score × 40% + Historical stability score × 30% + Carrier weight score × 30%, where, Real-time network metric score = 100 - (Packet loss rate × 50 + Latency × 0.2 + Jitter × 0.3).

[0098] The media stream is redirected to the network interface corresponding to the optimal path by calling the `bpf_redirect_map()` function through the eBPF program. During the redirection process, a double buffering mechanism is used to temporarily store data packets to ensure that the media stream is not lost, and the switching interruption time is less than 50ms. The primary hash table and the backup hash table are updated. The primary hash table stores the new four-tuple-path mapping, and the backup hash table caches historical mappings for rollback.

[0099] This invention achieves dynamic drift self-healing of media streams based on eBPF / XDP technology, and can repair NAT mapping failures without renegotiation.

[0100] S400 continuously monitors the signaling registration transaction logs on the business domain side through an independent trunk status maintenance module that operates independently of the internal business system. Based on the log cache, it caches the complete parameters of the most recent successful registration with the operator network. When the internal business system fails, it independently maintains the registration status and heartbeat with the operator network using the cached registration parameters to ensure the trunk link remains available.

[0101] In this embodiment, to address the problem of low reliability of relay links, i.e., when internal business system failures cause the operator-side link to be reset, the present invention uses "shadow online" technology and a hierarchical feedback mechanism to ensure that the relay link remains available when the internal business system fails.

[0102] This invention integrates signaling conversion and isolation, dynamic media stream optimization, and independent maintenance of relay status steps by adopting an asymmetric heterogeneous protocol stack architecture, thereby achieving highly secure, highly reliable, and highly compatible signaling conversion. It solves many defects of traditional signaling conversion systems, such as protocol attack risks, media stream interruption, single point of failure caused by service and relay coupling, and complex multi-operator adaptation.

[0103] In one embodiment, continuously monitoring the signaling registration transaction log on the business domain side includes capturing successful transactions of REGISTER requests in real time through at least one of the system log interface and a shared memory ring buffer.

[0104] In this embodiment, when the internal business system crashes, the relay registration status is lost, resulting in link interruption. This invention uses an independent agent to monitor registration transactions through the kernel log interface and shared memory ring buffer, and caches the most recently successfully registered parameters. In this way, the registration status can be maintained independently when the internal business system fails, avoiding link reset on the operator side and ensuring high availability of the relay link.

[0105] It should be noted that in existing technologies, a failure of the internal business system directly leads to the cessation of the heartbeat. This invention addresses this by having an independent module actively monitor the signaling logs (such as successful registration events) of the internal business system and cache key registration parameters in real time. In this way, even if the internal business system becomes completely unavailable, the independent module can still possess the "credential" of the last valid registration, thus fulfilling the prerequisite for maintaining the link state during a failure and achieving decoupling from the internal business system.

[0106] In one embodiment, it further includes:

[0107] When a fault is detected in an internal business system, corresponding actions are taken based on the fault severity level. The fault levels are categorized into Level 1, Level 2, and Level 3 faults, with Level 1 being less severe than Level 2, and Level 2 less severe than Level 3. The corresponding actions taken based on the fault severity level include:

[0108] When the fault severity level is Level 1, a temporary response indicating that the call has been queued is returned to the user terminal, and the call is queued to wait for the internal business system to be restored.

[0109] In the event of a level 2 fault, calls will be redirected to the emergency IVR module to provide basic voice services.

[0110] In the event of a level 3 fault, a pre-recorded fault alert will be played to the user, and the relay link will remain registered.

[0111] In this embodiment, a feedback mechanism is used to return a queued response, redirect to IVR, or play a prompt tone according to the fault level, ensuring that degraded services can still be provided during a fault, improving user experience, and guaranteeing basic communication capability grading. This can effectively solve the problem of user calls failing directly and resulting in a poor experience when internal business systems fail.

[0112] This embodiment provides service resilience assurance in the event of a failure. While independent modules maintain uninterrupted trunk links, a tiered feedback mechanism provides differentiated user handling strategies based on the severity of the internal business system failure. From simple queuing to providing emergency voice services and playing friendly prompts, it ensures that system behavior is controllable and orderly under any failure level, and provides appropriate feedback to users. This greatly enhances the end-user experience and the professionalism of the service, avoiding business losses and customer dissatisfaction caused by direct call failures.

[0113] Specifically, the registration transaction log listening unit listens to registered transactions on the business domain side in parallel using two methods:

[0114] Method 1: Capture all REGISTER request syslog events in real time through the Linux kernel log interface ( / dev / kmsg).

[0115] Method 2: Receive the registration success event pushed by the business domain protocol stack through a shared memory ring buffer (the / dev / shm area based on mmap mapping).

[0116] By deduplicating and verifying the consistency of captured events, we ensure that the same registered transaction is not recorded repeatedly and that the parameters are complete.

[0117] The registration parameter caching unit uses the LRU (Least Recently Used) algorithm to manage cache entries, caching the complete parameters of the most recently successfully registered entry (including Call-ID, Contact, Expires, etc.) to the in-memory database.

[0118] The fault detection unit monitors the operational status of internal business systems through a heartbeat mechanism. When three consecutive heartbeat packets (configurable) are lost, it is determined that the internal business system has failed.

[0119] The heartbeat maintenance unit immediately enables the cached registration parameters and sends a REGISTER heartbeat message (Expires=3600) to the IMS network in the exact same format as the last successful registration.

[0120] At the same time, the tiered feedback mechanism performs corresponding operations based on the fault level:

[0121] Level 1 Failure (Brief Internal Business System Jitter): A temporary response indicating that the call is queued is returned to the user terminal, waiting for recovery;

[0122] Level 2 Fault (Partial unavailability of internal business system functions): Redirect calls to the emergency IVR module;

[0123] Level 3 Fault (Complete Downtime of Internal Business Systems): Play a pre-recorded fault alert to the user and maintain the relay link registration status.

[0124] This invention uses an asymmetric heterogeneous protocol stack architecture and signaling desensitization mapping to physically isolate the service domain from the access domain protocol stack and block protocol-level attack paths, thus significantly improving the security of signaling conversion.

[0125] By leveraging eBPF / XDP technology to achieve dynamic drift self-healing at the kernel layer, media streams can be restored without SIP renegotiation in scenarios such as NAT drift, greatly enhancing media stream stability. Interruption time is less than 50ms, and the call interruption rate is reduced to below 0.1%. Through an independent trunk state maintenance module, registration with the operator network is maintained even when internal business systems fail, avoiding link resets and significantly improving trunk link reliability. IVR service interruption time is reduced from minutes to seconds. The pluggable protocol adaptation engine can dynamically load protocol completion plugins based on operator identifiers, automatically adapting to protocol differences between different operators. This shortens the development cycle for adding new operator adaptations and automates multi-operator compatibility.

[0126] like Figures 2 to 3 As shown, the present invention also provides a signaling conversion system based on telephone interaction, applied between the service domain and the operator access domain, comprising:

[0127] An asymmetric heterogeneous protocol stack architecture, including a business domain protocol stack and an access domain protocol stack;

[0128] The business domain protocol stack is deployed on the business domain side and is used to receive raw signaling messages from internal business systems;

[0129] The signaling desensitization mapping module is used to perform signaling desensitization mapping processing on the original signaling messages, remove the network topology sensitive fields, generate abstract metadata containing only business logic identifiers, and transmit the abstract metadata across domains to the access domain side through the kernel-level communication interface.

[0130] The access domain protocol stack is deployed on the access domain side and is used to convert abstract metadata into a carrier-recognizable signaling format and send it to the carrier network.

[0131] The media tunnel management module is used to establish media stream transmission tunnels at the operating system kernel layer, monitor the transmission quality indicators of the media stream transmission tunnels in real time, and trigger a dynamic drift self-healing process when the transmission quality indicators meet the preset degradation conditions, and switch the media stream to the optimized path without interrupting the existing media session. The transmission quality indicators include packet loss rate, latency and transmission quadruple.

[0132] The independent trunk status maintenance module operates independently of the internal business system. It continuously monitors the signaling registration transaction logs on the business domain side and caches the complete parameters of the most recent successful registration with the operator network based on the logs. When the internal business system fails, it uses the cached registration parameters to independently maintain the registration status and heartbeat with the operator network, so as to keep the trunk link available.

[0133] In this embodiment, the present invention integrates three core steps—signaling conversion and isolation, dynamic media stream optimization, and independent maintenance of relay state—by employing an asymmetric heterogeneous protocol stack architecture. This ensures high security, high reliability, and high compatibility in signaling conversion, thereby solving problems such as protocol attack risks, media stream interruption, single points of failure due to service and relay coupling, and complex multi-operator adaptation in traditional signaling conversion systems.

[0134] In this asymmetric heterogeneous protocol stack architecture, the service domain protocol stack adopts a lightweight instruction set (such as a simplified version of the SIP protocol), retaining only the necessary signaling fields; the access domain protocol stack fully implements the IMS standard protocol to ensure compatibility with operator equipment; among them, the signaling de-identification mapping converts SIP requests into abstract metadata through a preset template to avoid exposing protocol stack vulnerabilities.

[0135] The first signaling protocol stack (lightweight SIP) on the business domain side is deployed within the internal business system and only handles communication protocols within the business domain (such as the interaction between the bank's IVR system and the call center). When receiving raw SIP messages (such as REGISTER requests) from the internal business system, the signaling desensitization mapping module automatically removes sensitive fields (such as the IP address in the Via header and the terminal identifier in the Contact header), generates abstract metadata (such as Contact: sip:abstract_user@bank.com), and transmits it to the access domain side through the kernel-level communication interface (based on Linux's AF_NETLINK socket).

[0136] The access domain side second signaling protocol stack (complete IMS) is deployed on the operator's interface gateway, which converts abstract metadata into standard IMS signaling format (such as replacing the Contact field in the REGISTER request with an operator-recognizable URI) and interacts with the IMS network.

[0137] In the kernel-level logical port to media tunnel (i.e., media tunnel management module), media stream packets are intercepted at the kernel layer through eBPF / XDP technology, and the quadruple association matrix is ​​directly manipulated; the dynamic drift self-healing algorithm corrects the NAT mapping in real time through feature matching (such as SIPCall-ID) without triggering the Re-INVITE process.

[0138] In the independent relay state maintenance module, the IMS REGISTER state machine is maintained independently through a shadow online agent, ensuring registration even if the internal business system fails. A tiered feedback mechanism returns a temporary response to the user terminal indicating that the call is queued or redirects the call to the IVR module in the event of an internal business system failure. The shadow online agent monitors the registration transaction logs of the internal business system (such as syslog or a shared memory circular buffer), caching the complete parameters of the most recent successful REGISTER request in real time, thus maintaining the registration heartbeat after a service interruption.

[0139] It also employs a pluggable protocol compliance completion engine to automatically load protocol completion plugins based on the target gateway ID (such as operator identifier); and automatically repair missing SIP header fields (such as 100rel support) or adjust the interaction frequency (such as Session-Expires).

[0140] In the dynamic drift self-healing of eBPF / XDP tunnels, when the media stream transmission quality deteriorates (detection conditions include: packet loss rate > 10%, latency > 200ms, or source / destination quadruples abrupt change), the eBPF / XDP tunnel automatically triggers dynamic drift:

[0141] The XDP program resolves UDP quadruples (source IP, destination IP, source port, destination port) in real time. If a quadruple mutation is detected (such as a home router restart causing the source port to change from 5060 to 5061), it immediately switches to the backup tunnel path and updates the session context. The self-healing process takes less than 50ms.

[0142] By employing a shadow online agent's state synchronization mechanism, the shadow online agent caches registration parameters in real time based on the registration transaction log when the internal business system fails, ensuring the continuity of IMS registration. Specifically:

[0143] The shadow online agent listens for syslog events from the internal business system in real time through the Linux kernel log interface ( / dev / kmsg), capturing the complete parameters of all successful REGISTER requests (e.g., Contact: sip:user@bank.com;expires=3600 in REGISTER sip:bank.com SIP / 2.0).

[0144] Meanwhile, the agent synchronizes and registers transaction logs in real time through a shared memory ring buffer (based on the / dev / shm area mapped by mmap), avoiding data loss caused by log write delays;

[0145] When the internal business system is interrupted, the agent immediately activates the cached registration parameters and sends a REGISTER heartbeat message (Expires=3600) to the IMS network in the same format (e.g., Contact:sip:abstract_user@bank.com) to maintain the session state on the IMS side until the internal business system is restored.

[0146] This reduces IVR service interruption time from minutes to seconds (average <3 seconds in actual tests) during internal business system outages, thus avoiding link reset on the operator side.

[0147] This system can be deployed across cloud platforms without relying on physical leased lines, offering high architectural flexibility. Protocol stack isolation makes the core system completely transparent to external networks, reducing the attack surface by 90% and providing high hardware-level security. Through a dynamic drift self-healing algorithm, the media stream interruption rate is reduced to below 0.1%, achieving zero one-way call risk. When internal business systems restart, the operator-side trunk remains online, reducing maintenance costs by 70% and achieving high link robustness.

[0148] In one embodiment, the kernel-level media tunnel management module includes:

[0149] The quadruple association matrix manager is used to maintain the mapping relationship between media stream quadruples and transmission paths in real time.

[0150] The dynamic drift self-healing algorithm unit is used to calculate the optimal path and redirect the media stream when the transmission quality deteriorates;

[0151] The quality monitoring unit is used to collect data on packet loss rate, latency, and jitter in media streams in real time.

[0152] In this embodiment, since traditional media stream management requires a dedicated kernel-level module to achieve high performance and dynamic optimization, this system uses a kernel-level media tunnel management module, which includes a quadruple association matrix manager, a dynamic drift self-healing algorithm unit, and a quality monitoring unit, to achieve kernel-level media stream monitoring and dynamic optimization, resulting in high performance and low latency.

[0153] In one embodiment, the quadruple association matrix manager adopts a dual hash table structure and integrates a path prediction submodule; the path prediction submodule predicts sessions that may experience NAT drift within the next 5 minutes based on historical path switching records and real-time network topology, and pre-establishes backup tunnels.

[0154] In this embodiment, since passive switching can easily lead to temporary service degradation, this system integrates a path prediction submodule into the quadruple association matrix manager to predict NAT drift based on historical data and real-time topology, establishes backup tunnels in advance, and achieves predictive drift self-healing, further reducing the probability and duration of media stream interruptions.

[0155] In one embodiment, the independent relay status maintenance module includes:

[0156] Register a transaction log listening unit to support multi-source log collection;

[0157] Register a parameter caching unit and use the LRU algorithm to manage cache entries;

[0158] Heartbeat maintenance unit, supporting configurable heartbeat intervals and retry strategies;

[0159] The fault detection unit is used to detect the operating status of internal business systems and trigger hierarchical feedback.

[0160] In this embodiment, the independent relay status maintenance module includes log monitoring, parameter caching, heartbeat maintenance, and fault detection units, providing complete relay status maintenance functions to ensure rapid takeover and maintenance of the link in the event of a service failure.

[0161] In one embodiment, the pluggable protocol adaptation engine includes:

[0162] The protocol plugin registry center maintains a list of available plugins and their version information.

[0163] Plugin loader that supports hot-load and hot-unload protocol plugins;

[0164] The protocol compatibility testing unit is used to verify whether the signaling messages after the plugin modification comply with the target operator's specifications.

[0165] In this embodiment, since the plug-in protocol adaptation engine includes a protocol plugin registry, a plugin loader, and a protocol compatibility testing unit, it enables unified management and testing of plugins, ensuring the accuracy and reliability of protocol adaptation.

[0166] In one embodiment, the system is deployed on a dual-NIC server, wherein:

[0167] The first network card is a signaling network card, which is specifically used for signaling interaction with the operator's IMS network;

[0168] The second network interface card is a media network interface card, specifically used for media streaming transmission;

[0169] The eBPF / XDP program binds to two network cards respectively to achieve physical isolation between signaling and media streams.

[0170] In this embodiment, by deploying the system on a dual-NIC server, physically isolating the signaling NIC and the media NIC, and binding the eBPF / XDP program separately, physical isolation between signaling and media streams is achieved, avoiding interference and improving overall performance and security.

[0171] In one embodiment, the system further includes an AI-assisted decision-making module for:

[0172] Train a fault prediction model based on historical fault data to provide early warning of anomalies in internal business systems;

[0173] Optimize the path selection strategy for dynamic drift self-healing using reinforcement learning algorithms;

[0174] It automatically analyzes the protocol differences between different operators and generates new protocol completion plugin rules.

[0175] In this embodiment, the system announcement adds an AI-assisted decision-making module for fault prediction, path optimization, and protocol rule generation, thereby improving the system's adaptability, predictability, and optimization capabilities, reducing operation and maintenance costs, and avoiding the inability to adapt to dynamic and complex network environments due to reliance on static rules.

[0176] The AI-assisted decision-making module includes:

[0177] The fault prediction submodule is used to train an LSTM (Long Short-Term Memory) neural network model based on historical fault data (such as CPU utilization, memory leaks, and network jitter), and to analyze system monitoring indicators in real time to provide early warnings of internal business system anomalies.

[0178] The path optimization submodule is used to dynamically optimize the path scoring strategy using reinforcement learning algorithms (such as Deep Q-Network), adjusting the scoring weights based on real-time network conditions and service priorities (such as automatically selecting low-latency paths for VIP user calls).

[0179] The protocol rule generation submodule is used to automatically analyze the signaling interaction logs of different operators and extract protocol difference patterns.

[0180] This invention reduces the amount of manual writing work by automatically generating protocol completion plugin rule templates based on patterns.

[0181] It also includes an emergency IVR module; the emergency IVR module takes over calls when the internal business system fails, providing basic services such as voice navigation, recording, and call transfer. The emergency communication system includes:

[0182] The emergency IVR module is deployed on the access domain side to provide basic voice menus, recording, and call transfer functions. When a level 2 or 3 failure occurs in the internal business system, the call is automatically redirected to the emergency IVR module. The emergency IVR module can run independently, without relying on the internal business system database, and only provides predefined emergency service procedures.

[0183] In this embodiment, a highly available emergency communication system is built by having the IVR take over the calls in the event of a failure, so that voice services can still be provided when the internal business system fails.

[0184] The foregoing description of specific exemplary embodiments of the present invention is for illustrative and explanatory purposes. These descriptions are not intended to limit the invention to the precise forms disclosed, and it is obvious that many changes and variations can be made based on the above teachings. Although embodiments of the invention have been shown and described, these specific embodiments are merely explanations of the invention and are not intended to limit it. The specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. The purpose of selecting and describing exemplary embodiments is to explain the specific principles of the invention and its practical application, so that those skilled in the art, after reading this specification, can make modifications, substitutions, variations, and various choices and changes to the embodiments as needed without departing from the principles and spirit of the invention, provided that such modifications, substitutions, variations, and choices and changes are within the scope of the claims of the invention and are protected by patent law.

Claims

1. A signaling conversion method based on telephone interaction, applied between the service domain and the operator access domain, characterized in that, include: On the business domain side, the system receives raw signaling messages from internal business systems, performs signaling desensitization mapping processing on the raw signaling messages, removes network topology sensitive fields, generates abstract metadata containing only business logic identifiers, and transmits the abstract metadata across domains to the access domain side through a kernel-level communication interface. On the access domain side, the abstract metadata is converted into a carrier-recognizable signaling format and sent to the carrier network; A media stream transmission tunnel is established at the operating system kernel layer. The transmission quality indicators of the media stream transmission tunnel are monitored in real time. When the transmission quality indicators meet the preset degradation conditions, a dynamic drift self-healing process is triggered. The dynamic drift self-healing process includes redirecting the media stream to the optimal path calculated based on the quadrupole correlation matrix and path scoring algorithm through the eBPF program, so as to complete the switching without interrupting the existing media session. An independent relay status maintenance module, operating independently of the internal business system, continuously monitors the signaling registration transaction logs on the business domain side. Based on the log cache, it caches the complete parameters of the most recent successful registration with the operator network. When the internal business system fails, it independently maintains the registration status and heartbeat with the operator network using the cached registration parameters.

2. The signaling conversion method based on telephone interaction according to claim 1, characterized in that, Performing signaling desensitization mapping processing on the original signaling message includes: Sensitive fields in the original signaling message are identified using a preset sensitive field identification rule base; The abstract metadata is generated by replacing the specific values ​​of the sensitive fields with a preset mapping template. The sensitive fields include at least the IP address in the Via header, the terminal identifier in the Contact header, and the user identity information in the From / To header.

3. The signaling conversion method based on telephone interaction according to claim 2, characterized in that, The kernel-level communication interface is a socket based on the Linux kernel AF_NETLINK or AF_XDP, which is bound to the signaling network card and media network card of the dual network card server, and the signaling and media data are classified and isolated for transmission through the BPF program filter.

4. The signaling conversion method based on telephone interaction according to claim 1, characterized in that, The transmission quality metrics include packet loss rate, latency, and transmission quadruple, and the preset degradation conditions include at least one of the following: The average packet loss rate is greater than 10% over three consecutive sampling periods. The media stream has a one-way latency exceeding 200ms and lasts for more than 500ms. A sudden change occurs in any of the four elements of the media stream: source IP address, destination IP address, source port, and destination port.

5. The signaling conversion method based on telephone interaction according to claim 1, characterized in that, The eBPF program redirects media streams to the optimal path calculated using a quadruple association matrix and path scoring algorithm, including: The mapping relationship between media stream quadruples and transmission paths is maintained in real time through the quadruple association matrix manager; When transmission quality degradation is detected, the optimal path is calculated based on the current network topology and load status using a path scoring algorithm. The scoring factors of the path scoring algorithm include real-time network metrics, historical stability, operator priority, and cost weight. The real-time network metrics include the real-time packet loss rate, latency, and jitter of the path. The media stream is redirected to the optimal path via an eBPF procedure.

6. The signaling conversion method based on telephone interaction according to claim 5, characterized in that, The quadruple association matrix manager includes: The main hash table stores the currently active quadruple-path mappings and uses the RCU mechanism for lock-free updates. Backup hash table: used to cache historical four-tuple-path mappings and provide a fallback option in case of abnormal path switching; wherein, the update triggering conditions of the primary hash table and the backup hash table include at least one of NAT mapping changes, routing table updates and manual switching instructions.

7. The signaling conversion method based on telephone interaction according to claim 1, characterized in that, Continuously monitor the signaling registration transaction logs on the business domain side, including capturing successful transactions of REGISTER requests in real time through at least one of the following methods: system log interface and shared memory ring buffer.

8. The signaling conversion method based on telephone interaction according to claim 1, characterized in that, Also includes: When a fault is detected in the internal business system, corresponding operations are performed based on the fault severity level; wherein, the fault severity level includes Level 1 fault, Level 2 fault, and Level 3 fault, where Level 1 fault is less severe than Level 2 fault, and Level 2 fault is less severe than Level 3 fault. The corresponding operations performed based on the fault severity level include: When the severity level of the fault is Level 1, a temporary response indicating that the call has been queued is returned to the user terminal, and the call is queued to wait for the internal business system to be restored. If the fault severity level is level two, the call will be redirected to the emergency IVR module; When the fault severity level is three, a pre-recorded fault prompt tone is played to the user, and the relay link registration status is maintained.

9. The signaling conversion method based on telephone interaction according to claim 1, characterized in that, The step of converting signaling messages on the access domain side further includes a protocol compliance completion sub-step, which includes: Based on the target operator's identifier, dynamically load the corresponding protocol completion plugin from the protocol plugin library; The protocol completion plugin is used to perform compliance checks and corrections on the converted standard signaling messages; wherein the correction operations include at least one of adding missing SIP header fields, adjusting interaction frequency parameters, and correcting encoding formats.

10. A signaling conversion system based on telephone interaction, applied between the service domain and the operator access domain, characterized in that, include: An asymmetric heterogeneous protocol stack architecture, including a business domain protocol stack and an access domain protocol stack; The business domain protocol stack is deployed on the business domain side and is used to receive raw signaling messages from internal business systems; The signaling desensitization mapping module is used to perform signaling desensitization mapping processing on the original signaling message, remove the network topology sensitive fields, generate abstract metadata containing only business logic identifiers, and transmit the abstract metadata across domains to the access domain side through the kernel-level communication interface. The access domain protocol stack is deployed on the access domain side and is used to convert the abstract metadata into a signaling format that can be identified by the operator and send it to the operator's network. The media tunnel management module is used to establish a media stream transmission tunnel at the operating system kernel layer, monitor the transmission quality indicators of the media stream transmission tunnel in real time, and trigger a dynamic drift self-healing process when the transmission quality indicators meet the preset degradation conditions. The dynamic drift self-healing process includes redirecting the media stream to the optimal path calculated based on the quadrupole correlation matrix and path scoring algorithm through the eBPF program, so as to complete the switching without interrupting the existing media session. An independent relay status maintenance module operates independently of the internal business system. It continuously monitors the signaling registration transaction logs on the business domain side and caches the complete parameters of the most recent successful registration with the operator network based on the logs. When the internal business system fails, it independently maintains the registration status and heartbeat with the operator network using the cached registration parameters.

Images