System and methods for call Anti-spoofing in private branch exchange (PBX) systems

Abstract

A method for detection of a spoof call comprises: receiving a first voice response in response to a received call from a caller's device; augmenting the first voice response with a first watermark signal; sending the augmented first voice response to the caller's device; receiving a second voice response from the caller's device, wherein the second voice response is expected to contain therein an echo of the first watermark signal; analyzing the second voice response to determine the existence of the first watermark signal in the second voice response; and, generating a spoof call indication upon determination of the absence of the first watermark in the second voice response. System, circuit and non-transitory computer-readable medium embodiments are further provided.

Description

SYSTEM AND METHODS FOR CALL ANTI-SPOOFING IN PRIVATE BRANCH EXCHANGE (PBX) SYSTEMSCROSS REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Application No. 63 / 738,306 filed on December 23, 2024, the contents of which are incorporated herein by reference.TECHNICAL FIELD

[0002] The present disclosure generally relates to the field of anti-spoofing and particularly in anti-spoofing when calls are performed through a private branch exchange (PBX) to a call center.BACKGROUND

[0003] Private Branch Exchange (PBX) systems are integral to modern communication infrastructures, particularly in call centers where voice-based identification is commonly used. However, these systems are vulnerable to spoofing attacks, where an attacker mimics a legitimate user's voice to gain unauthorized access to sensitive information.

[0004] Automatic Speaker Verification (ASV) Systems are widely used for voice biometrics, providing a means to verify a speaker's identity based on their voice characteristics. These systems, however, are susceptible to various spoofing attacks, including synthetic speech, voice conversion, and replay attacks. To counter these threats, several advanced techniques have been developed.

[0005] Traditional methods rely on extracting specific features from the voice signal, such as Mel-frequency cepstral coefficients (MFCCs) and linear predictive coding (LPC). These features are then analyzed to detect anomalies indicative of spoofing. Recent advancements leverage deep learning models, such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), to automatically learn and detect spoofing patterns from raw audio data. These models have shown significant improvements in detecting sophisticated spoofing attempts.

[0006] End-to-end systems integrate the entire process of feature extraction, classification, and decision-making into a single model. This approach simplifies the detection pipeline and improves the robustness of anti-spoofing measures. For instance,CNN-GRU (Gated Recurrent Unit) architectures have been employed to enhance the detection of logical access attacks, such as speech synthesis and voice conversion.

[0007] Replay attacks involve using a recorded voice of a legitimate user to bypass security measures. To combat this, various techniques have been developed. For example, high-fidelity recording devices pose a significant threat, and countermeasures focus on detecting anomalies in the acoustic features of the recorded voice. Techniques such as phase information analysis and high-frequency noise detection are employed to identify replayed audio. Another approach may involve enhancing the robustness of ASV systems to different recording environments and devices is crucial. Methods like phase perturbation and channel compensation are used to improve the system's ability to distinguish between live and replayed audio.

[0008] Watermarking involves embedding a unique, imperceptible signal within the voice data that can be used to verify its authenticity. This technique is particularly effective against replay attacks and other forms of spoofing. For example, a watermark is embedded in the speech signal at the calling side and checked at the receiving side to verify the authenticity of the speaker. This method ensures that any tampering or replaying of the voice data can be detected by analyzing the watermark but requires that watermarking capabilities are present at the caller’s side, which is mostly impractical. Watermarks are designed to be robust against various types of attacks, including compression, noise addition, and filtering. This robustness ensures that the watermark remains detectable even if the voice data is manipulated.

[0009] A major challenge in anti-spoofing is the generalizability of detection systems across different datasets and real-world scenarios. Integrated solutions that combine multiple detection techniques and datasets have been proposed to address this issue. Cross-corpus evaluations are conducted to assess the effectiveness of these systems in diverse environments. Moreover, the advent of artificial intelligence (Al) solutions enable mimicking a caller’s voice from a snippet of the original voice making it more and more challenging to prevent such attacks, using the prior art. This is compounded with the use of vocoder distortions and low-quality compression, as well as the need for real-time cost effective solutions. Therefore, it is essential to provide a solution that overcomes the limitations and shortcomings of the prior art.SUMMARY

[0010] A summary of several example embodiments of the disclosure follows. This summary is provided for the convenience of the reader to provide a basic understanding of such embodiments and does not wholly define the breadth of the disclosure. This summary is not an extensive overview of all contemplated embodiments, and is intended to neither identify key or critical elements of all embodiments nor to delineate the scope of any or all aspects. Its sole purpose is to present some concepts of one or more embodiments in a simplified form as a prelude to the more detailed description that is presented later. For convenience, the term “some embodiments” or “certain embodiments” may be used herein to refer to a single embodiment or multiple embodiments of the disclosure.

[0011] Certain embodiments disclosed herein include a method for detection of a spoof call comprises: receiving a first voice response in response to a received call from a caller’s device; augmenting the first voice response with a first watermark signal; sending the augmented first voice response to the caller’s device; receiving a second voice response from the caller’s device, wherein the second voice response is expected to contain therein an echo of the first watermark signal; analyzing the second voice response to determine the existence of the first watermark signal in the second voice response; and, generating a spoof call indication upon determination of the absence of the first watermark in the second voice response.

[0012] Certain embodiments disclosed herein further include a system for enhancement of a private branch exchange comprises: a watermark generator; a mixer configured to mix a first watermark signal with a first voice signal resulting in an augmented voice signal that is sent over a voice network to a caller device; a voice activity detector (VAD) configured to receive a second voice signal and generate a voice detection signal responsive thereof; and, an echo analyzer configured to analyze the second voice signal upon receipt of an indication from the voice detection signal, wherein the analysis comprises causing the water mark generator to generate a first watermark signal to be mixed by the mixer.BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The subject matter disclosed herein is particularly pointed out and distinctly claimed in the claims at the conclusion of the specification. The foregoing and other objects, features, and advantages of the disclosed embodiments will be apparent from the following detailed description taken in conjunction with the accompanying drawings.

[0014] Figure 1 is a schematic communication diagram according to an embodiment.

[0015] Figure 2 is a schematic diagram of a communication system of a caller calling a call center through an enhanced private branch exchange (EPBX) according to an embodiment.

[0016] Figure 3 is a schematic diagram of communication system comprising a watermark generator and echo analyzer according to an embodiment.

[0017] Figure 4 is a schematic diagram of a portion of an enhancement circuit of an EPBX equipped with a watermark generator and echo analyzer according to an embodiment.

[0018] Figure 5 is a flowchart of operation of spoof detection according to an embodiment.DETAILED DESCRIPTION

[0019] It is important to note that the embodiments disclosed herein are only examples of the many advantageous uses of the innovative teachings herein. In general, statements made in the specification of the present application do not necessarily limit any of the various claims. Moreover, some statements may apply to some inventive features but not to others. In general, unless otherwise indicated, singular elements may be in plural and vice versa with no loss of generality. In the drawings, like numerals refer to like parts through several views.

[0020] A method for detection of a spoof call comprises: receiving a first voice response in response to a received call from a caller’s device; augmenting the first voice response with a first watermark; sending the augmented first voice response to the caller’s device; receiving a second voice response from the caller’s device, wherein the second voice response is expected to contain therein an echo of the first watermark; analyzing the second voice response to determine the existence of the first watermark in the second voice response; and, generating a spoof call indication upon determination of the absence of the first watermark in the second voice response.

[0021] Reference is now made to Fig. 1 depicting an example schematic communication diagram 100 according to an embodiment. Communication is shown to and from a caller’s device 110 with a communication network, for example a public switch telephone network (PSTN) 120. While a PSTN is described herein this should not be viewed as limiting on the invention and other voice carrying networks, for example, radio access network (RAN), are also possible separately or in any permissible combination. The caller’s device 120 is used by a caller 150 to communicate with one or more of call center personnel 160. The caller’s device 120 may be a cell phone, a smartphone, a landline telephone, and like voice communication devices. While reference is made herein to telephone calls, this should be interpreted broadly as implementable using devices other than telephones to perform a voice call. Over the PSTN 120 communication is directed to and from the caller device 110 and a telephone exchange, for example, an enhanced private branch exchange (EPBX) 130. The EPBX 130 is tasked with communication to and from a call center (CC) 140. The communication diagram 10 provides an explanation of the operation of the principles of the invention described herein.

[0022] A caller 150 initiates a call 110-1 to a CC 140 using a caller device 110. The call to CC 110-1 is routed 120-1 via the PSTN 120 to the EPBX 130, which is then further routed 130-1 to the CC 140. It is outside of the scope of the invention to discuss how the CC 140 reaches a specific CC person 160, however, it is well known in the art that the call may be answered by such a person at the CC 140 side. The connection signal 140- 1 is then routed from the CC 140 to the EPBX 130, which in turn routes 130-2 the connect call over the PSTN to be routed 120-2 to the caller’s device 110. As a result a call is established between the caller 150 and a responding person at the CC 140 side. It should be appreciated, that as automation grows, it’s possible that some, or all, of the communication described herein from the CC 140 side, is performed by synthesized voice, i,e., artificial voice generated by using one or more of a variety of technologies without departing from the scope of the invention.

[0023] As a call is established the caller 150 provides a caller voice 110-2 which is routed 110-2 to EPBX 120. EPBX 120 in turn routes 120-3 the caller voice to the EPBX 130, which in turn routes 130-3 to the CC 140. There, a CC voice response (human or artificial) is provided and routed 140-2 to the EPBX 130. The EPBX 130, adds a voice watermark130-5 to the CC voice response 130-5 routing them via PSTN 120. It should be understood that various techniques may be used to combine or mix, which may also be referred to as augmenting, the CC voice response 130-4 and the watermark 130-5 provided by the EPBX. The PSTN 120, routes the CC voice response 120-4 and the watermark 120-5 to the caller device 110.

[0024] The caller 150, using the caller’s device 110, responds with a caller voice 110-3 which is augmented by an echo of the watermark 110-4, and as further explained in respect of Fig. 3, both of which are routed 120-6 and 120-7 respectively by the PSTN 120 to the EPBX 130. The EPBX 120 performs an analysis of the signals received from the caller and identifies the echo of the watermark as matching the watermark sent by the EPBX 130. In an embodiment, if it is determined that there is no match with the watermark, and as further explained herein, then the EPBX 130 may send a spoof alert 130-6 to the CC 140. The EPBX 130 may further disconnect the call initiated by caller 150. There are various advantages over the prior art using the described communication diagram, especially because the watermark is independent of the caller device 110 and requires no configuration thereof.

[0025] In an embodiment the watermark is a unique watermark of the EPBX 130. In another embodiment, the watermark is generated periodically in order to prevent capturing of the watermark and attempting to spoof a call once such a watermark is detected. In yet another embodiment, a water mark is generated per call, i.e., the watermark is unique to each call, making it even more difficult to spoof a call. In yet another embodiment watermarks are sent periodically throughout a call interaction with the caller 150 to periodically determine the authenticity of the call. Moreover, this technique may augment other ant-spoofing techniques already used and provide further security.

[0026] Fig. 2 describes an example schematic diagram of a communication system 200 for a caller calling a call center 240 through an EPBX 230 according to an embodiment. The communication system 200 comprises a voice network 220, for example but not by way of limitation, a PSTN. The voice network 220 is communicatively connected to an EPBX 230, configured to operate as described herein, for example in Fig. 1. The voice network 220 is further communicatively connected, wired or wireless, to one or morecaller devices 210, for example call devices 210-1 through 210-n, where ‘n’ is an integer equal to or greater than ‘1 The EPBX 220 is connected to a CC 240 that comprises one or more telephone for receiving calls, for example, phones 240-1 through 240-m, where ‘m’ is an integer equal to or greater than ‘1

[0027] Operationally, each of the caller devices 210, the voice network 220, the EPBX 230, and the CC 240 operate as described with respects to call device 110, PSTN 120, EPBX 130 and call center 140, respectively in Fig. 1. That is, a caller may initiate a call using caller device 210-1 to be routed to the CC 240. Upon response from the CC 240 by a telephone, for example telephone 240-1 , a call is established as described with respect of Fig. 1. When a voice communication ensues, upon response from the telephone 240- 1 , the EPBX 230 may add, in any manner described herein, a watermark that is directed to the caller device 210-1. Upon listening to the call an echo of the watermark sent is captured by the caller device 210-1 , and as further explained herein and is directed back towards the EPBX 230 over the voice network 220. While CC 240 is described as comprising telephones manned by humans, this should not be viewed as limiting the generality of this invention. The CC 240 may comprises of bots that respond using any kind of synthesized voice, including the use of artificial intelligence (Al), to respond to a user. A CC voice response may be fully human, partly human and partly synthesized, or fully synthesized. The watermark, as explained herein, may be appended, added or mixed in a variety of ways, including those well known to those of skill in the art.

[0028] Reference is now made to Fig. 3 that depicts and example schematic diagram of communication system 300 comprising a watermark generator and echo analyzer (WMGEA) 390 according to an embodiment. A caller 310 performs a call using a caller device 320. The caller device comprises a microphone 326, a speaker 324, and an acoustic echo cancellation (AEG) unit 322. The AEG 322 is designed to prevent an acoustic echo through the communication loop from the microphone 326 to the speaker 324. One of skill in the art would appreciate that the acoustic echo cancellation is not perfect and depends on the characteristics of the audio received and the characteristics of the AEG 322. The caller device 320 is equivalent to the caller device 210 described in Fig. 2, and operative as described for caller device 110 in Fig. 1. Voice is communicated from 328 the caller device 320 to a voice network 330, for example but not by was oflimitation a PSTN, a RAN, or any permissible combination thereof, or vice versa to 334 the caller device 320.

[0029] According to the invention, an EPBX is enhanced by a WMGEA 390 that is communicatively connected to provide voice 382 to the voice network 330 or receive voice 332 from the voice network 330. The WMGEA 390 comprises a voice activity detector 340, an echo analyzer 350, a voice transmitter 360, a watermark generator 370, and a mixer 380. As a CO voice response, for example CO voice response 140-2, is provided 362 to the voice transmitter 360 it passes 364 to the mixer 380, and as may be appropriate, a watermark is appended, added, or mixed as provided from the watermark generator 370. The watermark is also provided to the echo analyzer 372 for analysis when the watermark echo is expected to be received. Under control 356 of the echo analyzer 350 an initial watermark is generated by watermark generator 370 if the VAD 340 detected voice activity in the received signal 332.

[0030] In an embodiment the watermark generator may provide a predetermined watermark, that may be hardwired or permanently programmed to the WMGEA 390. In another example, a random seed may be used to generate a unique watermark each time a watermark is to be used by the mixer 380. Using the echo loop, and as further discussed herein, the watermark generator 370 may regenerate additional watermark signals refining the signal to overcome the AEG 322 characteristics. This allows for detection of the watermark echo signal if such exists and make a determination regarding the authenticity of the caller 310. The number of iterations of generating a refined watermark signal may be predetermine for the WMGEA 390. The mixed voice signal 382 is sent via the voice network 330 to the caller device 320, which becomes a device under verification (DUV).

[0031] The voice and the watermark are received by the DUV 320, i.e.. the caller device, and the received voice+watermark is provided through the speaker 324 to 314 the caller 310 who can listen to it. However, this audio signal is also echoed 316 and captured by the microphone 326, which may also capture the voice of the caller 312. Now that echo with the traces of the watermark are sent back to the WMGEA 390 to be received by the VAD 340. The VAD 340 is designed to detect the presence of voice based on various known voice parameters. As noted, by fine tuning the watermark signal mixed with thevoice from the CC, it may sufficiently overcome the AEG 322 characteristic to allow the detection of the watermark at the EPBX side. If a spoof call is made then due to the characteristics of such attacks, the watermark signal will not be present raising the suspicion of the call being a spoof call. The voice signal itself is provided over 342 to the echo analyzer 342, as is, with a further detection signal 344 indicating if the received signal has voice characteristics or not.

[0032] The signals received by the echo analyzer 350 are the watermark 372, the detection signal 344 and the voice signal 342. The voice signal itself is provided to the CC, for example CC 240. The echo analyzer 350 further analyzes the traces of echo of the watermark and compares it, if present, to the received watermark from the watermark generator 370. The echo analyzer 350 is designed to provide a detection signal 352 indicating whether the caller is an illegitimate caller, or, if the call is legitimate as indicated on optional signal 354.

[0033] In an embodiment, the echo analyzer 350 may first operate to detect whether an acoustic echo of the watermark exists (‘yes’), does not exist (‘no’), or undetermined (‘sus[ect’). If ‘yes’, an acoustic reflection analysis may take place to determine whether the voice is a human voice, i.e., legitimate caller, or a synthesized voice, i.e., a spoof call. If the detection determines that no acoustic echo of the watermark exists, then the echo analyzer may generate an alert of a spoof call. In the case where it is not possible to determine a definite ‘yes’ or ‘no’, or if the number of iterations has not yet exceeded a predetermined limit, a fine tuning of parameters of the watermark signal is required and the echo analyzer 350 provides updated parameters for the updated watermark generation by watermark generator 370. Once a new watermark signal is generated it is mixed by mixer 380 and sent to the caller where the process is repeated and rechecked. This may repeat for a predetermined number of times before determining that is that there is no acoustic echo presenting the watermark. An alert signal 352 may be generated, as the call is considered suspect of being spoof call. While signals are described herein, other ways of indication may be used, including, but not by way of limitation, digital or analog messages, set or reset of indicator bits, and more, without departing from the scope of the invention.

[0034] In an embodiment, the echo analyzer 350, may further provide indications of failure to validate a received watermark if the characteristics of the watermark echo does not fit that of a reflection from a human. For example, a spoofer may wish to use a speaker to replay a that includes a watermark using a loudspeaker. The characteristics of a watermark echo when using a loudspeaker are different from those when the echo involved the caller 310 in the loop. Therefore, echo analyzer 350 is configured to detect such cases and when identified establishing the suspicion that a spoof call is attempted. That is, the echo analyzer 350 will determine the absence of a human caller from the loop raising such suspicion.

[0035] Fig. 4 is an example schematic diagram of an enhancement circuit 400 of an EPBX equipped with a watermark generator and echo analyzer according to an embodiment. A processing unit 420 is communicatively connected to a bus 410. The processing unit may comprise a central processing unit (CPU), a microprocessor, a digital signal processor (DSP), a microcontroller, and other like devices capable of processing instructions, one or more of same, as well as any permissible combinations thereof. The bus 410 may comprise a serial or parallel bus, have digital, analog, or mix thereof of communication capabilities, and designed to provide connectivity between components of the enhancement circuit 400.

[0036] The enhancement circuit 400 further comprises a memory 430 which may be a combination of random-access memory (RAM) and non-volatile memory (NVM) of a variety of types, which may further include a hierarch of memories such as primary and secondary caches, scratchpad memory, registers, solid-state disks (SSDs), and hard-disk drives (HDDs), and any permissible combination thereof. A portion of memory 430 may be dedicated for code memory 435. Code memory 435 contains therein, either permanently or temporarily instructions that when executed by the processing unit 420 provide a private branch exchange the certain enhancements that allow the EPBX to perform the teachings herein.

[0037] An input / output (IO) interface 440 provides for external signal to communicatively connect to the enhancement circuit 400 and such connectivity may include various kinds of networking capabilities such as, but not limited to, universal serial bus (USB), ethernet, Bluetooth®, near-field communication (NFC), and other relevant networks, whether wiredor wireless. The IO interface 440 is communicatively connected to the bus 410 for the purposes of sending and receiving signals and data as the case may be.

[0038] A watermark generator and echo analyzer (WMGEA) 450 is also communicatively connected to the bus 450. WMGEA 450 may further have external communication for receiving certain control and data signals, separate from the IO interface 440. However, in an embodiment such communication, in full or in part, may be available using the IO interface 440. The functionality of the WMGEA 450 has been described in detail with respect of WMGEA 390. It should be understood that certain elements of the operation of WMGEA 390 or WMGEA 400 may be performed at least partially by execution of instruction stored in memory 430 and executed by the processing unit 420.

[0039] Reference is now made to Fig. 5 which depicts an example flowchart 500 of operation of spoof detection according to an embodiment. In an embodiment of the flowchart 500, at least S530 through S590 may be implemented using WMGEA 400.

[0040] At S510 a call is received from a caller, for example caller 150 shown in Fig. 1.

[0041] At S520 it is checked whether the call has been answered by CO, for example, CO 140 shown in Fig. 1. If the call was answered then execution continues with S530; otherwise, execution completes.

[0042] At S530, a watermark signal is generated and is then appended, added, or mixed with a voice response provided by the CO. In an embodiment this may be done conditional upon first receiving a voice indication from the caller to the CO. The generation of the watermark signal may initially be, for example but not by way of limitation, a redetermined watermark signal or a randomly generated watermark signal. In repetitions of the loop S530 through S570, the watermark signal may be refined based on watermark echo signal to allow to overcome the acoustic echo cancellation mechanisms typically residing on a caller’s device, and as further explained herein.

[0043] At S540 the voice+watermark are sent back to the caller, so that the caller may listen to the message using a speaker of a caller’s device, for example, the caller device 110 shown in Fig. 1 and microphone 326 shown in Fig. 3.

[0044] At S550 a response is received from the caller which may include an echo of the watermark sent and replayed on the caller’s device and received by a microphone of the caller device, for example, microphone 326.

[0045] At S560 the received voice and watermark are analyzed to determine the existence of a watermark.

[0046] At S570 it is checked whether the caller is a valid caller based on the analysis made at S560. In an embodiment, if the caller is valid then execution continues with S580; if the caller is not valid, execution continues with S590; and, if the analysis is inconclusive, the call is suspect and another watermark iteration takes place and therefore execution continues with S530. The number of iterations on an inconclusive result may be limited to a predetermined number of validation attempts before a decision is taken, by for example WMGEA 400, to either identify the caller as valid or invalid, i.e., continue with S580 or S590 respectively.

[0047] At S580 a valid signal is provided to indicate that the caller was validated. In an embodiment this is an optional signal which may not be implemented.Jn yet another embodiment the valid signal is also accompanied by a confidence level indication so that a decision can be made as to additional measures, if any, need to be taken in this case.

[0048] At S590 an indication of a spoof call is provided. In an embodiment the invalid signal is also accompanied by a confidence level indication so that a decision can be made as to additional measures, if any, need to be taken in this case.

[0049] All examples and conditional language recited herein are intended for pedagogical purposes to aid the reader in understanding the principles of the disclosed embodiment and the concepts contributed by the inventor to furthering the art and are to be construed as being without limitation to such specifically recited examples and conditions. Moreover, all statements herein reciting principles, aspects, and embodiments of the disclosed embodiments, as well as specific examples thereof, are intended to encompass both structural and functional equivalents thereof. Additionally, it is intended that such equivalents include both currently known equivalents as well as equivalents developed in the future, i.e., any elements developed that perform the same function, regardless of structure.

[0050] It should be understood that any reference to an element herein using a designation such as “first,” “second,” and so forth does not generally limit the quantity or order of those elements. Rather, these designations are generally used herein as a convenient method of distinguishing between two or more elements or instances of anelement. Thus, a reference to first and second elements does not mean that only two elements may be employed there or that the first element must precede the second element in some manner. Also, unless stated otherwise, a set of elements comprises one or more elements.

[0051] As used herein, the phrase “at least one of” followed by a listing of items means that any of the listed items can be utilized individually, or any combination of two or more of the listed items can be utilized. For example, if a system is described as including “at least one of A, B, and C,” the system can include A alone; B alone; C alone; 2A; 2B; 2C; 3A; A and B in combination; B and C in combination; A and C in combination; A, B, and C in combination; 2A and C in combination; A, 3B, and 2C in combination; and the like.

Claims

CLAIMSWhat is claimed herein is exemplary and non-limiting on the scope of the invention:

1. A method for detection of a spoof call comprises:receiving a first voice response in response to a received call from a caller’s device;augmenting the first voice response with a first watermark signal;sending the augmented first voice response to the caller’s device;receiving a second voice response from the caller’s device, wherein the second voice response is expected to contain therein an echo of the first watermark signal;analyzing the second voice response to determine the existence of the echo of the first watermark signal in the second voice response, wherein the echo of the first watermark has traversed the caller device’s acoustic echo cancellation circuitry; and generating a spoof call indication upon determination of the absence of the first watermark in the second voice response.

2. The method of claim 1 , wherein the watermark is one of: a permanent watermark and a unique watermark.

3. The method of claim 1 , further comprising:generating a valid signal upon determination of the existence of the first watermark in the second voice response.

4. The method of claim 1 , wherein the analysis further determines an inability to determine if the caller is valid or invalid.

5. The method of claim 4, further comprising:receiving a third voice response in response to a received call from a caller’s device;augmenting the third voice response with a second watermark signal, wherein the second watermark signal is refined to overcome acoustic echo cancelation at the caller’s side;sending the augmented third voice response to the caller’s device;receiving a fourth voice response from the caller’s device, wherein the fourth voice response is expected to contain therein an echo of the second watermark signal;analyzing the fourth voice response to determine the existence of the second watermark signal in the fourth voice response; andgenerating an invalid signal upon determination of the absence of the second watermark in the fourth voice response.

6. The method of claim 1 , wherein generating a spoof call indication further comprises determination of a spoof call when the watermark echo characteristics are indicative of absence of a caller from the loop.

7. The method of claim 1 , wherein the spoof call indication is at least one of: a signal, a digital message, an analog message.

8. A system for enhancement of a private branch exchange comprises:a watermark generator;a mixer configured to mix a first watermark signal with a first voice signal resulting in an augmented voice signal that is sent over a voice network to a caller device;a voice activity detector (VAD) configured to receive a second voice signal and generate a voice detection signal responsive thereof; andan echo analyzer configured to analyze the second voice signal upon receipt of an indication from the voice detection signal, wherein the analysis comprises causing the water mark generator to generate a first watermark signal to be mixed by the mixer.

9. The system of claim 8, wherein the echo analyzer is further configured to extract an echo of the watermark signal returned from the caller device through the voice network to the system, further configured to determine if the echo of the watermark signal is of the first watermark signal, and further configured to assert a spoof call indication upon invalidity determination.

10. The system of claim 9, wherein the spoof call indication is at least one of: a signal, a digital message, an analog message.

11. A non-transitory computer-readable medium storing instructions which, when executed by processing circuitry of an enhanced private branch exchange (EPBX), cause the EPBX to perform operations comprising:receiving a first voice response from a caller device over a voice network; generating a first watermark signal and causing the first watermark signal to be mixed with the first voice response to create an augmented voice response;transmitting the augmented voice response toward the caller device such that audio comprising the augmented voice response is emitted through a speaker of the caller device and at least a portion thereof is acoustically re-captured by a microphone of the caller device;receiving a second voice response from the caller device, the second voice response including audio generated by the caller and an acoustic echo of the first watermark signal, subject to acoustic echo cancellation processes on the caller device;analyzing the second voice response to detect whether the acoustic echo of the first watermark signal is present therein; andasserting a spoof-call indication when the analysis determines an absence of the acoustic echo of the first watermark signal.

12. The non-transitory computer-readable medium of claim 11 , wherein the instructions further cause the EPBX to:generate an updated watermark signal refined to compensate for acoustic echo cancellation characteristics of the caller device upon an inconclusive determination of the presence of the acoustic echo; andrepeat the transmitting and analyzing operations using the updated watermark signal.

13. The non-transitory computer-readable medium of claim 11 , wherein the watermark generator is further instructed to produce a watermark signal that is one of: unique to the call, periodically changed during the call, or randomly generated per transmission iteration.

14. The non-transitory computer-readable medium of claim 11 , wherein the analyzing further comprises:identifying acoustic characteristics indicative of replay through a loudspeaker device and asserting the spoof-call indication when the acoustic echo exhibits loudspeaker-origin signatures instead of human-coupled acoustic reflections.

15. The non-transitory computer-readable medium of claim 11 , wherein asserting the spoof-call indication further comprises:generating a confidence level associated with the determination to enable adaptive call-handling policies.

16. An enhancement circuit for a private branch exchange (PBX), the enhancement circuit comprising:a processing unit communicatively coupled to a bus;a memory storing instructions that, when executed by the processing unit, cause the enhancement circuit to:receive a first voice response from a caller device over a voice network; cause a watermark generator to generate a watermark signal;cause a mixer to mix the watermark signal with a voice signal originating from a call center to produce an augmented voice signal for transmission toward the caller device;receive a second voice response from the caller device, the second voice response including audio produced by the caller device after emission of the augmented voice signal through a speaker of the caller device;analyze the second voice response, using an echo analyzer, to determine whether the second voice response includes an acoustic echo of the watermark signal; and,assert a spoof-call indication when the acoustic echo of the watermark signal is absent.

17. The enhancement circuit of claim 16, wherein the instructions further cause the enhancement circuit to:refine one or more parameters of the watermark signal to compensate for acoustic echo cancellation characteristics of the caller device, upon an inconclusive determination of the presence of the acoustic echo of the watermark signal; and, repeat the mixing, transmitting, and analyzing operations using the refined watermark signal.

18. The enhancement circuit of claim 16, wherein the watermark generator is further configured to generate watermark signals that are one of: unique per call, periodically updated during the call, or randomly generated for each iteration of the spoof-detection process.

19. The enhancement circuit of claim 12, wherein analyzing the second voice response further comprises:determining whether the acoustic echo exhibits acoustic characteristics indicative of replay through a loudspeaker device instead of a human-coupled acoustic reflection; and,asserting the spoof-call indication when loudspeaker-origin characteristics are detected.

20. The enhancement circuit of claim 12, wherein the watermark generator is activated to generate the watermark signal responsive to a voice detection indication produced by the voice activity detector.

Images