A one-way flow table telemetry system and method
By generating a joint signature of compressed digest and leakage limit proof, combined with time window watermarking, the problems of unquantifiable information leakage and untrustworthy equipment in one-way transmission devices are solved, achieving efficient information leakage detection and reliable equipment operation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG RUIWEN TECH CO LTD
- Filing Date
- 2025-10-24
- Publication Date
- 2026-07-31
AI Technical Summary
Existing one-way transmission devices lack the measurability and provability of the amount of leaked information, cannot verify the reliable operation of the device in an isolated state, and lack a secure compression and verification mechanism for cross-time-series flow table statistical data.
Design a one-way flow table telemetry system. The sending end generates a compressed digest M(t) and a leakage limit proof P(t), and performs joint signing using the device's private key. The receiving end verifies the integrity and authenticity of the data frame's source, and uses a time window watermark W(t) to verify and archive the amount of information leakage.
It enables measurable and verifiable information leakage, ensures the reliable operation of devices in an isolated state, supports high throughput and low overhead for millions of flow tables, and can detect scanning attacks and DDoS attacks.
Smart Images
Figure CN121567362B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of network communication technology, and in particular to a one-way flow meter telemetry system and method. Background Technology
[0002] Unidirectional transmission devices (network gateways) are used to ensure one-way communication between different security domains and are commonly found in industrial control, classified, and financial networks. Against the backdrop of rapid digitalization and networking, national critical information infrastructures are facing higher demands for network isolation and data security.
[0003] Traditional network gateway devices are characterized by physical unidirectional transmission, typically using fiber optic cables or FPGA transfer units to achieve unidirectional data exchange, but they have the following shortcomings:
[0004] (1) Lack of measurability and verifiability of the amount of leaked information;
[0005] (2) Unable to verify the trusted operation of the device in an isolated state;
[0006] (3) Lack of a secure compression and verification mechanism for cross-time-series flow table statistics.
[0007] Existing technology 1, CN202321598385.7, mainly focuses on "high-speed data synchronization structure" to achieve high-bandwidth data transfer through FPGA channels, but does not involve information leakage constraints or verifiable proof.
[0008] For example, existing technology 2: CN202311366698.4 involves "cross-class signature transmission", focusing on the electronic signature verification process. Summary of the Invention
[0009] This invention addresses the problems in existing one-way transmission devices, such as the lack of measurability and verifiability of the amount of leaked information, the inability to verify the reliable operation of the device in an isolated state, and the lack of a secure compression and verification mechanism for cross-time-series flow table statistical data. It provides a one-way flow table telemetry system and method.
[0010] To solve the above-mentioned technical problems, the present invention provides the following technical solution:
[0011] A one-way flow table telemetry system includes a transmitter and a receiver. The transmitter is deployed in a first security domain and is configured to: generate a compressed digest M(t) from flow table hit data within a statistical time window; and generate a leakage upper limit proof P(t) based on the compressed digest M(t).
[0012] A time-window watermark W(t) is generated based on a policy ratchet token; a device trust certificate Q(t) is generated; the time-window watermark W(t) is injected into the compressed digest M(t); the compressed digest M(t), the leakage limit proof P(t), and the device trust certificate Q(t) are jointly signed using the device private key to generate device-level signature data Sig2; the sending end sends a telemetry data frame containing M(t), P(t), Q(t), and Sig2 via a unidirectional physical link. The telemetry data frame includes the compressed digest M(t), the leakage limit proof P(t), the device trust certificate Q(t), and the device-level signature data Sig2; the receiving end is deployed in the second security domain and is configured to: receive the telemetry data frame; verify the validity of the certificate chain in the device trust certificate Q(t); and verify the device-level signature data Sig2 using the public key corresponding to the device identity in the device trust certificate Q(t) to confirm the integrity and authenticity of the source of the compressed digest M(t), the leakage limit proof P(t), and the device trust certificate Q(t). The leakage limit proof P(t) verifies that the information leakage of the compressed digest M(t) does not exceed the preset threshold; after the verification is passed, the compressed digest M(t) is archived or subjected to situational awareness analysis.
[0013] Preferably, the compressed summary M(t) generated from the flow table hit data within the statistical time window includes:
[0014] The set of statistical flow tables within the time window is F. Perform hit statistics to obtain a hit count set. : ,
[0015] Where: f i For the i-th flow table entry, C i The number of times a flow table entry is hit within a time window; n is the number of valid flow tables within that time window.
[0016] Obtain the compressed digest M(t) from the hit count set Ct: M(t) = Compress(Ct);
[0017] Where: Compress() is a compression function that performs spatial compression and sensitivity desensitization on the flow table statistics; M(t) is the compressed summary.
[0018] As a preferred embodiment, a leakage limit proof P(t) is generated based on the compressed digest M(t), and signature data Sig1 is generated to bind the compressed digest M(t) and the leakage proof P(t) to prevent tampering.
[0019] Using information entropy H(M(t)) and statistical distance Generate proof of the leakage upper bound P(t), ;
[0020] Where H(M(t)) is the information entropy of the compressed digest M(t); The statistical distance between the output distribution differences under different sensitivity states;
[0021] ;
[0022] Where p(x) is the frequency of occurrence of symbol or pattern x in the compressed digest M(t); The maximum permissible entropy threshold is set.
[0023] ;
[0024] in, , These represent the output distributions of the system under two desensitization states; The maximum allowable distribution difference;
[0025] The signature data Sig1 is generated by binding the compressed digest M(t) with the leakage upper limit proof P(t). ;
[0026] Here, Hash() is a secure hash function used to generate digest signatures, || is a concatenation operation, and Sig1 is signature data used to bind compressed digests and leak proofs to prevent tampering.
[0027] As a preferred approach, the time window watermark W(t) is generated based on a strategy ratchet token, for the pseudo-random watermark seed. The strategy ratchet token of the i-th window of the strategy ratchet. Determine the pseudo-random watermark seed with window number i , ;
[0028] Where H() is a one-way hash function, Let i be the policy ratchet token for the i-th window, where i is the window number;
[0029] Using pseudo-random watermark seed i Generate a time window watermark W(t).
[0030] As a preferred approach, the time window watermark W(t) is injected into the compressed digest M(t); the compressed digest M(t), the leakage limit proof P(t), and the device trust proof Q(t) are jointly signed using the device private key to generate device-level signature data Sig2. ;
[0031] Where Q(t) is the device trust proof; For the device's private key;
[0032] Q(t)=(DevID, CertChain, hash(bitstream||policy), window_id, monotonic_ctr);
[0033] Among them, DevID is the unique identifier of the device; CertChain is the certificate chain used to verify the legitimacy of DevID and the signing public key; Bitstream is the bit stream or firmware image of the programmable logic in the left field; policy is the version, whitelist or configuration of the metric-related policy; window_id is the time window number; and monotonic_ctr is a monotonically increasing counter that resists replay.
[0034] Preferably, the receiver is also configured to independently reconstruct the desired time window watermark W'(t) based on the locally generated policy ratchet token and the current time window identifier, and to verify whether the received compressed digest M(t) contains a watermark trace that matches the desired time window watermark W'(t).
[0035] Preferably, the unidirectional physical link includes a rate shaping module and a jitter template.
[0036] As a preferred option, both the transmitter and receiver are implemented using FPGAs and processed using a pipelined architecture.
[0037] To address the aforementioned technical problems, this invention also provides a one-way flow table telemetry method, comprising a transmitter and a receiver. The transmitter is deployed in a first security domain, and the receiver is deployed in a second security domain, connected via a one-way physical link. The method includes:
[0038] The steps executed by the sending end:
[0039] The flow table hit data within the statistical time window is used to generate a compressed summary M(t);
[0040] Generate the leakage upper limit proof P(t) based on the compressed digest M(t);
[0041] Generate a device trust proof Q(t);
[0042] A time window watermark W(t) is generated based on a policy ratchet token, and the time window watermark W(t) is injected into the compressed digest M(t);
[0043] The device private key is used to jointly sign the compressed digest M(t), the proof of leakage limit P(t), and the proof of device trust Q(t) to generate device-level signature data Sig2;
[0044] Telemetry data frames are sent via a one-way physical link. The telemetry data frames contain a compressed digest M(t), a leakage limit proof P(t), a device trust proof Q(t), and device-level signature data Sig2.
[0045] The steps performed by the receiving end are as follows:
[0046] Receive telemetry data frames;
[0047] Verify the validity of the certificate chain in the device trust proof Q(t);
[0048] The device-level signature data Sig2 is verified using the public key corresponding to the device identity in the device trust proof Q(t) to confirm the integrity and authenticity of the source of the compressed digest M(t), the leakage limit proof P(t), and the device trust proof Q(t);
[0049] Based on the leakage limit proof P(t), it is verified that the information leakage of the compressed digest M(t) does not exceed the preset threshold.
[0050] After both verification and validation are passed, the compressed summary M(t) is archived or used for situational awareness analysis.
[0051] Preferably, the receiving end also includes independently reconstructing the desired time window watermark W'(t) based on the locally generated policy ratchet token and the current time window identifier, and verifying whether the received compressed digest M(t) contains a watermark trace that matches the desired time window watermark W'(t).
[0052] This invention, by adopting the above technical solutions, has significant technical effects:
[0053] The unidirectional flow meter telemetry system designed in this invention fixes "who / what / when" through equipment measurement proof (Q(t)), and its authenticity can be traced.
[0054] This invention introduces the information theory leakage upper bound proof into a network gateway. The leakage upper bound proof P(t) sets an upper bound on the output information quantity using entropy and statistical distance; and calculates the entropy constraint and statistical distance in real time at the nanosecond level through an FPGA hardware pipeline.
[0055] The strategy of this invention allows the ratchet to operate continuously without backtracking, and the hardware hash pipeline generates tokens that can only evolve forward, thus avoiding replay / rollback attacks.
[0056] The link shape constraint + thresholding digest of this invention has strong resistance to covert channels.
[0057] This invention generates M(t) / P(t) through hardware pipeline, supports millions of flow tables, and features high throughput and low overhead.
[0058] This invention can indicate a scanning attack when the abstract shows an abnormal increase in the active flow table cardinality and a sparse distribution;
[0059] When the digest shows that the number of hits in some flows is much higher than the average, it can indicate a DDoS attack. These detections rely only on digest characteristics and do not involve the specific communication content, yet they still meet the constraints of the proof of the upper limit of information leakage. Attached Figure Description
[0060] Figure 1 This is a schematic diagram of the system of the present invention.
[0061] Figure 2 This is a schematic diagram of the transmitting end of the present invention.
[0062] Figure 3 This is a schematic diagram of the receiving end of the present invention. Detailed Implementation
[0063] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.
[0064] Example 1
[0065] A one-way flow meter telemetry system includes a transmitter and a receiver.
[0066] The sender is deployed in the first security domain and is configured to: generate a compressed digest M(t) based on the flow table hit data within the statistical time window; and generate a leak limit proof P(t) based on the compressed digest M(t).
[0067] A time-window watermark W(t) is generated based on a policy ratchet token; a device trust certificate Q(t) is generated; the time-window watermark W(t) is injected into the compressed digest M(t); the compressed digest M(t), the leakage limit proof P(t), and the device trust certificate Q(t) are jointly signed using the device private key to generate device-level signature data Sig2; the sending end sends a telemetry data frame containing M(t), P(t), Q(t), and Sig2 via a unidirectional physical link. The telemetry data frame includes the compressed digest M(t), the leakage limit proof P(t), the device trust certificate Q(t), and the device-level signature data Sig2; the receiving end is deployed in the second security domain and is configured to: receive the telemetry data frame; verify the validity of the certificate chain in the device trust certificate Q(t); and verify the device-level signature data Sig2 using the public key corresponding to the device identity in the device trust certificate Q(t) to confirm the integrity and authenticity of the source of the compressed digest M(t), the leakage limit proof P(t), and the device trust certificate Q(t). The leakage limit proof P(t) verifies that the information leakage of the compressed digest M(t) does not exceed the preset threshold; after the verification is passed, the compressed digest M(t) is archived or subjected to situational awareness analysis.
[0068] This embodiment is implemented based on Xilinx Ultrascale+ FPGA and uses a unidirectional fiber optic physical link.
[0069] Nominal rate of unidirectional link: 10 Gbps; Effective rate: 9.8 ± 0.1 Gbps after rate shaping and jitter template constraint; Time window length: 1 s; Number of statistical entries per window: 10 6 End-to-end delay: ≤ 1 window;
[0070] Signature algorithm: SM2, single signature latency ≤ 120 µs;
[0071] The compressed summary M(t) generated from the flow table hit data within the statistical time window includes:
[0072] The set of statistical flow tables within the time window is F. Perform hit statistics to obtain the hit count set C. t : ,
[0073] Where: f i For the i-th flow table entry; C i The number of hits for the flow table entry within the time window; n is the number of valid flow tables within the time window;
[0074] Obtain the compressed digest M(t) from the hit count set Ct: M(t) = Compress(Ct);
[0075] Where: Compress() is a compression function that performs spatial compression and sensitivity desensitization on the flow table statistics; M(t) is the compressed summary.
[0076] The compressed structure uses a Bitmap binary value to indicate whether a flow table entry is active (hit count > 0), used for activity detection; CMS (Count-MinSketch) frequency estimation approximates the traffic frequency, providing statistical information in the frequency dimension; HLL (HyperLogLog) cardinality estimation is used to estimate the cardinality of the flow table set (number of active flow tables); the above structures are combined, i.e.: M(t) = Bitmap(F)||CMS(F)||HLL(F), where F is the traffic data.
[0077] Hardware pipelines can perform field-level anonymization, including masking, hashing, and regular expression matching, to ensure that specific identification information is not exposed during the compression stage, and only statistical features are retained.
[0078] Based on the compressed digest M(t), a leakage limit proof P(t) is generated, and signature data Sig1 is generated to bind the compressed digest M(t) and the leakage proof P(t) to prevent tampering.
[0079] Using information entropy H(M(t)) and statistical distance Generate proof of the leakage upper bound P(t), ;
[0080] Where H(M(t)) is the information entropy of the compressed digest M(t); The statistical distance between the output distribution differences under different sensitivity states;
[0081] ;
[0082] Where p(x) is the frequency of occurrence of symbol or pattern x in the compressed digest; The maximum permissible entropy threshold is set.
[0083] ;
[0084] in, , These represent the output distributions of the system under two desensitization states; The maximum allowable distribution difference; when The smaller the value, the less likely an attacker is to distinguish the system state based on the output, and the lower the risk of leakage.
[0085] The signature data Sig1 is generated by binding the compressed digest M(t) with the leakage upper limit proof P(t). ;
[0086] Here, Hash() is a secure hash function used to generate digest signatures, || is a concatenation operation, and Sig1 is signature data used to bind compressed digests and leak proofs to prevent tampering.
[0087] The time window watermark W(t) is generated based on the strategy ratchet token, and the pseudo-random watermark seed is... i The strategy ratchet token is used through the i-th window of the strategy ratchet. i The pseudo-random watermark seed is determined by the window number i. i , ;
[0088] Where H() is a one-way hash function, and Token i Let i be the policy ratchet token for the i-th window, where i is the window number;
[0089] Using pseudo-random watermark seed i Generate a time window watermark W(t).
[0090] Inject the time window watermark W(t) into the compressed digest M(t); use the device private key to jointly sign the compressed digest M(t), the leakage limit proof P(t), and the device trust proof Q(t) to generate device-level signature data Sig2. ;
[0091] Where Q(t) is the device trust proof; For the device's private key;
[0092] Q(t)=(DevID, CertChain, hash(bitstream||policy), window_id, monotonic_ctr);
[0093] Among them, DevID is the unique identifier of the device; CertChain is the certificate chain used to verify the legitimacy of DevID and the signing public key; Bitstream is the bit stream or firmware image of the programmable logic in the left field; policy is the version, whitelist or configuration of the metric-related policy; window_id is the time window number; and monotonic_ctr is a monotonically increasing counter that resists replay.
[0094] The receiver is also configured to independently reconstruct the desired time window watermark W'(t) based on the locally generated policy ratchet token and the current time window identifier, and to verify whether the received compressed digest M(t) contains a watermark trace that matches the desired time window watermark W'(t).
[0095] The unidirectional physical link includes a rate shaping module and a jitter template. The template parameters are: a rate of 9.8 ± 0.1 Gbps and a jitter ratio of no more than 1%. Both the transmitter and receiver are implemented using FPGAs and processed in a pipelined architecture.
[0096] Example 2
[0097] Based on Example 1, this example is a one-way flow table telemetry method, including a sending end and a receiving end. The sending end is deployed in a first security domain, and the receiving end is deployed in a second security domain, connected via a one-way physical link. The method includes:
[0098] The steps executed by the sending end:
[0099] The flow table hit data within the statistical time window is used to generate a compressed summary M(t);
[0100] Generate the leakage upper limit proof P(t) based on the compressed digest M(t);
[0101] Generate a device trust certificate Q(t); generate a time window watermark W(t) based on a policy ratchet token, and inject the time window watermark W(t) into the compressed digest M(t); use the device private key to jointly sign the compressed digest M(t), the leakage limit certificate P(t), and the device trust certificate Q(t) to generate device-level signature data Sig2;
[0102] Telemetry data frames are sent via a one-way physical link. The telemetry data frames contain a compressed digest M(t), a leakage limit proof P(t), a device trust proof Q(t), and device-level signature data Sig2.
[0103] The steps performed by the receiving end are as follows:
[0104] Receive telemetry data frames; verify the validity of the certificate chain in the device's trusted certificate Q(t);
[0105] The device-level signature data Sig2 is verified using the public key corresponding to the device identity in the device trust proof Q(t) to confirm the integrity and authenticity of the source of the compressed digest M(t), the leakage limit proof P(t), and the device trust proof Q(t);
[0106] Based on the leakage limit proof P(t), it is verified that the information leakage of the compressed digest M(t) does not exceed the preset threshold.
[0107] After both verification and validation are passed, the compressed summary M(t) is archived or used for situational awareness analysis.
[0108] The receiving end also includes independently reconstructing the desired time window watermark W'(t) based on the locally generated policy ratchet token and the current time window identifier, and verifying whether the received compressed digest M(t) contains a watermark trace that matches the desired time window watermark W'(t).
[0109] Introducing predictable self-test beacons in the absence of a reverse channel for detecting missed detections / tampering:
[0110] The time window watermark W(t) is generated based on the strategy ratchet token, and the pseudo-random watermark seed is... i The strategy ratchet token is used through the i-th window of the strategy ratchet. i The pseudo-random watermark seed is determined by the window number i. i ;
[0111] Where H() is a one-way hash function, and Token i Let i be the policy ratchet token for the i-th window, where i is the window number;
[0112] Using pseudo-random watermark seed i Generate a time window watermark W(t);
[0113] The left-side smart gateway card is based on the seed i Inject controllable weak beacons (such as micro-increments at a fixed number of bucket / bit positions or specific register bit patterns) into the statistical domain, and the receiver reconstructs the strategy ratchet token using the same i-th window. i Using i, calculate the pseudo-random watermark seed. i The receiver independently reconstructs the desired time window watermark W'(t); the receiver determines whether the received compressed digest M(t) contains traces of the expected time window watermark W(t); if the missing / deviation exceeds the threshold, it is deemed suspicious (missed report / forgery). The time window watermark W(t) does not carry sensitive business information, only verifiable structure, and is still subject to the leakage limit proof P(t).
[0114] The policy ratchet ensures that the policy can proceed safely without rollback when there is no reverse link, and can be verified by the right-side receiver, as shown in Table 1.
[0115] Table 1 shows the strategy ratchet and time window.
[0116]
[0117] Pre-generate strategies for N future time windows {S1, ..., S2} N},calculate:
[0118] Where K is the pre-shared key / facility-built-in key; Token i When the window token time advances to window i, the policy ratchet token for window i.
[0119] The left-side sender retrieves S from the local policy table. i Locally compute the policy ratchet token for the i-th window. i The right-side receiver controls K and S. i The generation rules, locally calculate the policy ratchet token for the i-th window. i Consistent time progression, i.e., ratchet-based advancement, eliminates the need for a reverse link.
[0120] This embodiment is implemented on an FPGA and uses a pipeline structure to handle flow table statistics and summary compression. It processes no less than 10 data points per time window. 6The flow table entries have a bucket count m ≥ 4096, a Count-Min Sketch depth d = 4, a width w = 2048, an HLL error range of 0.5%-2%, an upper limit of θ entropy of 10-12 bits, an upper limit of ε distance of 0.02-0.05, and can use SM2 / ECC for encryption signatures and SM3 for hashing. Table 2 shows the resource usage of the main FPGA modules (taking Xilinx Ultrascale+ series as an example).
[0121] Table 2 Resource Usage Table for Main FPGA Modules
[0122]
[0123] Example 3
[0124] Based on the above embodiments, this embodiment includes the following step: Step 1, the left-side sending end uses a Bitmap to mark whether a million-level flow table entry has been hit; calculate H(B) ≤ 500 bits; where Bitmap is a compressed data structure that uses binary bits (0 or 1) to represent whether each flow table entry has been hit (i.e., there is a traffic match) within a time window. For example, if there are 1 million flow table entries, Bitmap is a 1 million-bit sequence, where 1 represents active (hit count > 0) and 0 represents inactive. This method is efficient and space-saving, but only retains the active state and does not expose specific traffic details. H(B) is the information entropy of Bitmap (in bits), which measures the amount of information carried by Bitmap. The higher the entropy value, the greater the risk of information leakage.
[0125] Step 2: Generate a leak proof P(t) to prevent tampering with the signature data Sig1; use information entropy H(M(t)) and statistical distance... Generate proof of the leakage upper bound P(t), ;
[0126] Where H(M(t)) is the information entropy of the compressed digest M(t); The statistical distance between the output distribution differences under different sensitivity states;
[0127] ;
[0128] Where p(x) is the frequency of occurrence of symbol or pattern x in the compressed digest M(t); The maximum permissible entropy threshold is set.
[0129] ;
[0130] in, , These represent the output distributions of the system under two desensitization states; The maximum allowable distribution difference; when The smaller the value, the less likely an attacker is to distinguish the system state based on the output, and the lower the risk of leakage.
[0131] The signature data Sig1 is generated by binding the compressed digest M(t) with the leakage upper limit proof P(t). ;
[0132] Here, Hash() is a secure hash function used to generate digest signatures, || is a concatenation operation, and Sig1 is signature data used to bind compressed digests and leak proofs to prevent tampering.
[0133] Step 3: Generate Q(t) from device metrics and bitstream hash, and sign it Sig2; generate a time window watermark W(t) based on the policy ratchet token, and use a pseudo-random watermark seed. i The strategy ratchet token is used through the i-th window of the strategy ratchet. i The pseudo-random watermark seed is determined by the window number i. i ;
[0134] Where H() is a one-way hash function, and Token i Let i be the policy ratchet token for the i-th window, where i is the window number;
[0135] Using pseudo-random watermark seed i Generate a time window watermark W(t).
[0136] Inject the time window watermark W(t) into the compressed digest M(t); use the device private key to jointly sign the compressed digest M(t), the leakage limit proof P(t), and the device trust proof Q(t) to generate device-level signature data Sig2. ;
[0137] Where Q(t) is the device trust proof; For the device's private key;
[0138] Q(t)=(DevID, CertChain, hash(bitstream||policy), window_id, monotonic_ctr);
[0139] Among them, DevID is the unique identifier of the device; CertChain is the certificate chain used to verify the legitimacy of DevID and the signing public key; Bitstream is the bit stream or firmware image of the programmable logic in the left field; policy is the version, whitelist or configuration of the metric-related policy; window_id is the time window number; and monotonic_ctr is a monotonically increasing counter that resists replay.
[0140] By binding Q(t), M(t), and P(t) together, we can prove that the data source is authentic and generated by a trusted device.
[0141] One-way transmission {M(t), P(t), Q(t), W(t), Sig1, Sig2};
[0142] Right-side receiver verification: CertChain / Sig2
[0143] Use a certificate chain to verify the validity of the DevID and public key (check the validity period, revocation list, etc.).
[0144] Verify Sig2 with a public key to ensure that Q(t), M(t), and P(t) have not been tampered with and come from a trusted device.
[0145] Verify Sig1: Verify Sig1 to ensure that M(t) and P(t) are correctly bound.
[0146] Verify H and Δ thresholds: Check if the entropy H(M(t)) and statistical distance Δ in P(t) meet preset thresholds (e.g., H ≤ θ, Δ ≤ ε). If they exceed these thresholds, it indicates a high risk of information leakage, and the data may be discarded. Verify the right side of W(t) using a local token. i Calculate seed_i with i to reconstruct the expected watermark. Compare the received M(t) to see if it contains the correct watermark pattern. If the deviation exceeds a threshold, it is considered suspicious (possibly due to missed detection or forgery). Verify link shape: Check if the received data rate, frame length, etc., conform to the preset template to suppress timing covert channels. Post-verification audit / situational archiving: After all verifications pass, the data is used for security auditing and network situational analysis (such as detecting scanning attacks or DDoS trends).
[0147] Example 4
[0148] In this embodiment, the left-side transmitter of CMS+HLL+P(t)+Q(t)+W(t) (scale and anomaly) uses CMS to statistically analyze the frequency and estimate the base; calculates Δ≤0.05 and H(M(t))≤θ, and generates P(t) and Sig_1; generates device trust proof Q(t), device-level signature data Sig2, and injects time window watermark W(t);
[0149] The right side confirms the trustworthiness of the executable environment and bitstream through device trustworthiness proof Q(t), and combines the leakage upper limit proof P(t) threshold and time window watermark W(t) watermark determination to determine effectiveness; used to identify the trend and scale of anomalies such as DDoS scanning (without exposing specific connections).
[0150] Example 5
[0151] This embodiment uses a strategy ratchet + shape template (unattended operation); the left-side sending end has a pre-set 12-month strategy S1…S 12 Calculate the policy ratchet token for the i-th window. i ; A 128B digest frame is sent every 10ms; The right-side receiver only verifies the Token, shape template, and time window watermark W(t); Zero reverse link evolution strategy throughout the year; Alarm is triggered if there is an abnormal shape / watermark mismatch.
Claims
1. A one-way flow meter telemetry system, comprising a transmitter and a receiver, characterized in that, The sender is deployed in the first security domain and is configured as follows: The flow table hit data within the statistical time window is used to generate a compressed summary M(t); Generate the leakage upper limit proof P(t) based on the compressed digest M(t); A time window watermark W(t) is generated based on a strategy-based ratchet token. Generate a device trust proof Q(t); Inject the time window watermark W(t) into the compressed digest M(t); use the device private key to jointly sign the compressed digest M(t), the leakage limit proof P(t), and the device trust proof Q(t) to generate device-level signature data Sig2; The sending end transmits a telemetry data frame containing M(t), P(t), Q(t) and Sig2 via a one-way physical link. The telemetry data frame includes a compressed digest M(t), a leak limit proof P(t), a device trust proof Q(t), and device-level signature data Sig2. The receiver is deployed in the second security domain and is configured as follows: Receive telemetry data frames; Verify the validity of the certificate chain in the device trust proof Q(t); The device-level signature data Sig2 is verified using the public key corresponding to the device identity in the device trust proof Q(t) to confirm the integrity and authenticity of the compressed digest M(t), the leak limit proof P(t), and the device trust proof Q(t); Based on the leakage limit proof P(t), it is verified that the information leakage of the compressed digest M(t) does not exceed the preset threshold. After the verification is passed, the compressed summary M(t) is archived or used for situational awareness analysis.
2. The one-way flow table telemetry system of claim 1, wherein, The compressed summary M(t) generated from the flow table hit data within the statistical time window includes: a set of flow tables F within a time window, ; perform hit statistics to obtain a set of hit counts : ; Where: f i For the i-th flow table entry; C i For this flow table entry in the time window The number of hits within the time window; n is the number of valid flow tables within that time window; Obtain the compressed digest M(t) from the hit count set Ct: M(t) = Compress(Ct); Where: Compress() is a compression function that performs spatial compression and sensitivity desensitization on the flow table statistics; M(t) is the compressed summary.
3. The one-way flow table telemetry system of claim 1, wherein, A leakage upper bound proof P(t) is generated based on the compressed digest M(t), and a signature data that binds the compressed digest M(t) and the leakage proof P(t) to prevent tampering is generated ; By information entropy H(M(t)) and statistical distance Generating a leakage upper bound proof P(t), ; H(M(t)) is the information entropy of the compressed digest M(t) after compression; is the statistical distance of the output distribution difference under different sensitivity states; ； where p(x) is the frequency of occurrence of the symbol or pattern x in the compressed summary M(t); is a set maximum allowed entropy threshold value; ； wherein, , are the output distributions of the system in the two desensitization states, respectively; is the maximum allowed distribution difference. The signature data Sig1 is generated by binding the compressed digest M(t) with the leakage upper bound proof P(t), ; Here, Hash() is a secure hash function used to generate digest signatures, || is a concatenation operation, and Sig1 is signature data used to bind compressed digests and leak proofs to prevent tampering.
4. The one-way flow table telemetry system of claim 1, wherein, Based on the strategy of generating time window watermarks W(t) using ratchet tokens, and for pseudo-random watermark seed... i The strategy ratchet token is passed through the i-th window of the strategy ratchet. i The pseudo-random watermark seed is determined by the window number i. i , ; Where H() is a one-way hash function, and Token i Let i be the policy ratchet token for the i-th window, where i is the window number; by a pseudo-random watermark seed seed i generating a time window watermark W(t).
5. The one-way flow table telemetry system of claim 2, wherein, Inject the time window watermark W(t) into the compressed digest M(t); use the device private key to jointly sign the compressed digest M(t), the leakage limit proof P(t), and the device trust proof Q(t) to generate device-level signature data Sig2. ; Wherein, Q(t) is the device trust proof; is the device private key; Q(t)=(DevID, CertChain, hash(bitstream||policy), window_id, monotonic_ctr); Among them, DevID is the unique identifier of the device; CertChain is the certificate chain used to verify the legitimacy of DevID and the signing public key; Bitstream is the bit stream or firmware image of the programmable logic in the left field; policy is the version, whitelist or configuration of the metric-related policy; window_id is the time window number; and monotonic_ctr is a monotonically increasing counter that resists replay.
6. The one-way flow table telemetry system of claim 1, wherein, The receiver is also configured to independently reconstruct the desired time window watermark W'(t) based on the locally generated policy ratchet token and the current time window identifier, and to verify whether the received compressed digest M(t) contains a watermark trace that matches the desired time window watermark W'(t).
7. The one-way flow table telemetry system of claim 1, wherein, A unidirectional physical link includes a rate shaping module and a jitter template.
8. The one-way flow table telemetry system of claim 1, wherein, Both the transmitter and receiver are implemented using FPGAs and processed in a pipelined architecture.
9. A one-way flow table telemetry method, comprising a transmitter and a receiver, wherein the transmitter is deployed in a first security domain and the receiver is deployed in a second security domain, and the two are connected via a one-way physical link, the method comprising: The steps executed by the sending end: The flow table hit data within the statistical time window is used to generate a compressed summary M(t); Generate the leakage upper limit proof P(t) based on the compressed digest M(t); Generate a device trust proof Q(t); A time window watermark W(t) is generated based on a policy ratchet token, and the time window watermark W(t) is injected into the compressed digest M(t); The device private key is used to jointly sign the compressed digest M(t), the proof of leakage limit P(t), and the proof of device trust Q(t) to generate device-level signature data Sig2; Telemetry data frames are sent via a one-way physical link. The telemetry data frames contain a compressed digest M(t), a leakage limit proof P(t), a device trust proof Q(t), and device-level signature data Sig2. The steps performed by the receiving end are as follows: Receive telemetry data frames; Verify the validity of the certificate chain in the device trust proof Q(t); The device-level signature data Sig2 is verified using the public key corresponding to the device identity in the device trust proof Q(t) to confirm the integrity and authenticity of the compressed digest M(t), the leakage limit proof P(t), and the device trust proof Q(t); Based on the leakage limit proof P(t), it is verified that the information leakage of the compressed digest M(t) does not exceed the preset threshold. After both verification and validation are passed, the compressed summary M(t) is archived or used for situational awareness analysis.
10. A unidirectional flow meter telemetry method according to claim 9, characterized in that, The receiving end also includes independently reconstructing the desired time window watermark W'(t) based on the locally generated policy ratchet token and the current time window identifier, and verifying whether the received compressed digest M(t) contains a watermark trace that matches the desired time window watermark W'(t).