1<!DOCTYPE html> 2<html> 3 <head> 4 <link rel="icon" href="http://example.com/favicon.png"> 5 <title>Shravya Kanchi</title> 6 <meta charset="utf-8"> 7 <meta name="viewport" content="width=device-width, initial-scale=1"> 8
8<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.6.4/jquery.min.js"></script>
8 9
9<script src="https://maxcdn.bootstrapcdn.com/bootstrap/3.4.1/js/bootstrap.min.js"></script>
9 10 <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.0.0-beta3/css/all.min.css"> 11 <link rel="stylesheet" href="https://maxcdn.bootstrapcdn.com/bootstrap/3.4.1/css/bootstrap.min.css"> 12 <link rel="stylesheet" href="https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700&display=swap"> 13 <style> 14 body { 15 position: relative; 16 background-color: #f2e4d6; 17 font-family: "Inter", system-ui, -apple-system, BlinkMacSystemFont, sans-serif; 18 font-size: 16px; 19 line-height: 1.65; 20 letter-spacing: 0.01em; 21 font-weight: 400; 22 color: #111827; 23 } 24 25 .timeline { 26 border-left: 1px solid hsl(0, 0%, 90%); 27 position: relative; 28 list-style: none; 29 } 30 31 .timeline .timeline-item { 32 position: relative; 33 } 34 35 .timeline .timeline-item:after { 36 position: absolute; 37 display: block; 38 top: 0; 39 } 40 41 .timeline .timeline-item:after { 42 background-color: hsl(0, 0%, 90%); 43 left: -46px; 44 border-radius: 50%; 45 height: 11px; 46 width: 11px; 47 content: ""; 48 } 49 50 #profile { 51 padding-top: 50px; 52 } 53 54 #section0 { 55 padding-top: 50px; 56 color: #000000; 57 } 58 59 #section1 { 60 color: #000000; 61 } 62 63 #section2 { 64 color: #000000; 65 } 66 67 #section3 { 68 color: #000000; 69 } 70 71 #section4 { 72 color: #000000; 73 } 74 75 #section5 { 76 color: #000000; 77 } 78 79 .video-grid { 80 display: grid; 81 grid-template-columns: repeat(2, 1fr); 82 gap: 16px; 83 } 84 .panel-info { 85 border: 1px solid rgba(0, 0, 0, 0.06); 86 border-radius: 10px; 87 box-shadow: 0 1px 6px rgba(0, 0, 0, 0.06); 88 margin-bottom: 28px; 89 } 90 91 .video-grid-item { 92 overflow: hidden; 93 border-radius: 8px; 94 box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1); 95 width: 100%; 96 padding-bottom: 75%; 97 /* Set to maintain a 1:1 aspect ratio */ 98 position: relative; 99 } 100 101 .video-grid-item iframe, 102 .video-grid-item img { 103 width: 100%; 104 height: 100%; 105 object-fit: cover; 106 position: absolute; 107 top: 0; 108 left: 0; 109 } 110 111 .carousel { 112 width: auto; 113 /* Set the width of the carousel */ 114 margin: auto; 115 /* Center the carousel */ 116 } 117 118 .carousel-inner .item { 119 height: auto; 120 /* Set the height of the carousel items */ 121 } 122 123 .embed-responsive { 124 padding-bottom: 56.25%; 125 /* 16:9 aspect ratio */ 126 height: 0; 127 overflow: hidden; 128 max-width: 100%; 129 background: #000; 130 } 131 132 .embed-responsive iframe { 133 position: absolute; 134 top: 0; 135 left: 0; 136 width: 100%; 137 height: 100%; 138 } 139 .navbar-custom { 140 background-color: #f2e4d6; 141 } 142 .navbar-custom .nav > li > a { 143 padding: 16px 18px; 144 font-weight: 500; 145 } 146 .navbar-custom .nav > li.active > a, 147 .navbar-custom .nav > li > a:focus, 148 .navbar-custom .nav > li > a:hover { 149 font-weight: 600; 150 text-decoration: underline; 151 text-underline-offset: 4px; 152 } 153 154 .well { 155 background-color: #fff7ef; 156 border-color: #e3d3c2; 157 color: #2b2420; 158 } 159 160 body, h1, h2, h3, h4, h5, h6, p, a, li { 161 color: #111827; 162 } 163 164 .section-title { 165 font-size: 28px; 166 font-weight: 600; 167 letter-spacing: -0.02em; 168 margin-bottom: 1.4rem; 169 color: #2E4057; 170 } 171 172 h4 { 173 font-size: 20px; 174 font-weight: 600; 175 letter-spacing: -0.01em; 176 margin-top: 2.2rem; 177 margin-bottom: 0.4rem; 178 } 179 180 .project-title { 181 font-size: 18px; 182 font-weight: 600; 183 letter-spacing: -0.01em; 184 margin-top: 2rem; 185 margin-bottom: 0.6rem; 186 } 187 .project-title:first-of-type { 188 margin-top: 0.3rem; 189 } 190 191 .project-label { 192 font-weight: 600; 193 } 194 195 ul li { 196 font-size: 16px; 197 font-weight: 400; 198 letter-spacing: 0.01em; 199 margin-bottom: 0.45rem; 200 } 201 202 .meta { 203 font-size: 14px; 204 color: #6b7280; 205 letter-spacing: 0.01em; 206 } 207 208 .section-spacing { 209 margin-bottom: 34px; 210 } 211 212 .panel-body {
213 padding: 22px 26px; 214 } 215 216 .intro-lead h4 { 217 font-size: 20px; 218 font-weight: 600; 219 letter-spacing: -0.01em; 220 } 221 222 .intro-lead p { 223 font-size: 18px; 224 letter-spacing: 0.01em; 225 } 226 .intro-lead p { 227 letter-spacing: 0.005em; 228 } 229 230 .project-block { 231 margin-top: 18px; 232 padding-top: 8px; 233 border-top: 1px dashed rgba(0, 0, 0, 0.08); 234 } 235 236 .project { 237 margin-bottom: 2.6rem; 238 } 239 240 .project-line { 241 margin-bottom: 0.35rem; 242 } 243 244 h4 + p, 245 h4 + ul { 246 margin-top: 0.6rem; 247 } 248 249 .education-item { 250 margin-bottom: 20px; 251 } 252 253 .edu-row { 254 display: flex; 255 justify-content: space-between; 256 gap: 12px; 257 flex-wrap: wrap; 258 } 259 260 .edu-degree { 261 font-weight: 600; 262 } 263 264 .edu-years, 265 .edu-cgpa { 266 color: #6b7280; 267 font-size: 14px; 268 letter-spacing: 0.01em; 269 } 270 271 .edu-cgpa { 272 margin-top: 4px; 273 font-size: 16px; 274 } 275 276 .skills-title { 277 font-size: 18px; 278 font-weight: 600; 279 letter-spacing: -0.01em; 280 margin-top: 1.2rem; 281 margin-bottom: 0.5rem; 282 } 283 284 #section1 em, 285 #section2 em, 286 #section3 em, 287 #section4 em, 288 #section5 em, 289 #section6 em { 290 color: #6b7280; 291 font-size: 14px; 292 font-style: normal; 293 font-weight: 500; 294 letter-spacing: 0.01em; 295 } 296 297 @media (max-width: 991px) { 298 .col-sm-3, 299 .col-sm-9 { 300 padding-left: 15px !important; 301 padding-right: 15px !important; 302 } 303 304 .panel-body { 305 padding: 18px 18px; 306 } 307 308 #profile { 309 padding-top: 20px; 310 } 311 } 312 </style> 313 </head> 314 <body data-spy="scroll" data-target=".myNavbar" data-offset="50"> 315
315<script src="https://code.jquery.com/jquery-3.5.1.slim.min.js"></script>
315 316
316<script src="https://cdn.jsdelivr.net/npm/@popperjs/[email protected]/dist/umd/popper.min.js"></script>
316 317
317<script src="https://stackpath.bootstrapcdn.com/bootstrap/4.5.2/js/bootstrap.min.js"></script>
317 318 <nav class="navbar navbar-expand-lg navbar-custom"> 319 <div class="container-fluid"> 320 <div class="navbar-header"> 321 <button type="button" class="navbar-toggle" data-toggle="collapse" data-target="#myNavbar"> 322 <span class="icon-bar"></span> 323 <span class="icon-bar"></span> 324 <span class="icon-bar"></span> 325 </button> 326 <a class="navbar-brand" href="#"></a> 327 </div> 328 <div class="collapse navbar-collapse" id="myNavbar"> 329 <ul class="nav navbar-nav navbar-right"> 330 <li class="active"> 331 <a href="#about">About</a> 332 </li> 333 <li> 334 <a href="#research">Research</a> 335 </li> 336 <li> 337 <a href="#publications">Publications</a> 338 </li> 339 <li> 340 <a href="#section6">Experience</a> 341 </li> 342 <li> 343 <a href="./Resume_Shravya_2026.pdf" target="_blank">Resume</a> 344 </li></ul> 345 </div> 346 </div> 347 </nav> 348 <div class="row"> 349 <div class="col-sm-3" style=" padding-right: 0px;padding-left: 0px;"> 350 <!-- <div class="affix"> --> 351 <div id="profile" class="container-fluid" style="padding-top: 0px;padding-bottom: 10px; padding-left: 20px; padding-right: 10px;"> 352 <img src="pic-2026.jpg" class="img-responsive img-rounded center-block" style=".img-fluid. max-width: auto; height:auto;"> 353 </div> 354 <!-- <h2 class = "text-center"><span class="label label-success">Shravya Kanchi</span></h2> --> 355 <div class="well well-lg"> 356 <h2 style="margin:10px;padding-left: 20px;padding-top: 0px; padding-bottom: 10px;padding-right: 10px;">Shravya Kanchi</h2> 357 <h4 style="margin:10px;padding-left: 20px;padding-right: 10px;">PhD, AI Security Researcher</h4> 358 <h4 style="margin:10px;padding-left: 20px;padding-right: 10px;">Computer Science Department, Virginia Tech</h4> 359 <h4 style="margin:10px;padding-left: 20px;padding-right: 10px;"> 360 <a href="mailto: [email protected]"><i class="fas fa-envelope"></i> [email protected] </a> 361 </h4> 362 <h4 style="margin:10px;padding-left: 20px;padding-right: 10px;"> 363 <a href="https://scholar.google.co.in/citations?user=VsdYF68AAAAJ&hl=en/" target="_blank"> 364 <i class="fas fa-graduation-cap"></i> Google Scholar </a> 365 </h4> 366 <h4 style="margin:10px;padding-left: 20px;padding-right: 10px;"> 367 <a href="https://github.com/shrave" target="_blank"> 368 <i class="fab fa-github"></i> GitHub </a> 369 </h4> 370 <h4 style="margin:10px;padding-left: 20px;padding-right: 10px;"> 371 <a href="https://www.linkedin.com/in/shravya-kanchi/" target="_blank"> 372 <i class="fab fa-linkedin"></i></i> LinkedIn </a> 373 </h4> 374 <h4 style="margin:10px;padding-left: 20px;padding-right: 10px;"> 375 <a href="./Resume_Shravya_2026.pdf" target="_blank"> 376 <i class="fas fa-file-alt"></i> Resume </a> 377 </h4> 378 </div> 379 <!-- </div> --> 380 </div> 381 <div class="col-sm-9" style="padding-top: 0px; padding-right: 100px; padding-left: 0px;"> 382 <div id="section0" class="container-fluid"> 383 <div id="about"></div> 384 <!-- <h3 style="color: #ffffff;"> Welcome</h3> --> 385 <div class="" style="margin-right:20px; text-align: left;"> 386 <div class="panel-body intro-lead"> 387 <p> 388 389 <h4>AI Security Researcher</h4> 390 <p>I recently earned my PhD in Computer Science from Virginia Tech, Blacksburg, where I was advised by <a href="https://people.cs.vt.edu/~danfeng/" target="_blank"><strong>Dr. Daphne Yao</strong></a>. I work at the intersection of <strong>security and generative AI</strong>, turning advanced AI methods into practical defenses for real-world cybersecurity problems. My research spans LLM safety, synthetic data generation, and adversarial ML, with publications in <strong>IEEE S&P, Asia CCS, and ACSAC</strong>. I build evaluation-driven systems that improve security robustness in practice, drawing on hands-on experience with LLMs, GANs, and diffusion models for security-specific data generation and safer chatbot customization.</p> 391 <p>I am currently seeking <strong>industry research</strong> or <strong>postdoctoral</strong> roles starting in Summer 2026 and welcome inquiries via email.</p> 392 393 <!-- <p><b>I plan to spend Summer 2025 working on impactful research at the intersection of Security and Generative AI. If you have relevant opportunities, please reach out to me on LinkedIn or via email.</b></p> --> 394 </div> 395 </div> 396 </div> 397 <div id="section1" class="container-fluid section-spacing"> 398 <div id="updates"></div> 399 <h3 class="section-title">Updates</h3> 400 <div class="panel panel-info" style="margin-right:20px"> 401 <div class="panel-body"> 402 <div class="row"> 403 <div class="col-md-12"> 404 <!-- <ul> --> 405 <!-- NEW UPDATE START --> 406 <div> 407 <em>May 2026: </em> I successfully defended my PhD dissertation, <strong>âGenerative AI for Hardening Security: From Training Data Augmentation, Language Alignment, to Code Vulnerability Demonstration.â</strong> 408 </div> 409 <div> 410 <em>May 2026: </em> Our new arXiv preprint <a href="https://arxiv.org/abs/2605.03956" target="_blank"><strong>âGenerating Proof-of-Vulnerability Tests to Help Enhance the Security of Complex S
410oftwareâ</strong></a> is now available. 411 </div> 412 <div> 413 <em>Feb 2026: </em> Guest Lecturer for <strong>CS 5590 (Systems and Software Security)</strong>, presenting a talk on <strong>âLLMs for Supply Chain Security.â</strong> 414 </div> 415 <div> 416 <em>Feb 2026: </em> Our paper <a href="https://arxiv.org/abs/2507.05660" target="_blank"><strong>âOptimus: A Robust Defense Framework for Mitigating Toxicity while Fine-Tuning Conversational AIâ</strong></a> was accepted to <a href="https://www.codaspy.org/2026/" target="_blank"><strong>ACM CODASPY 2026</strong></a>. 417 </div> 418 <div> 419 <em>Dec 2025: </em> Invited to serve as a guest judge for the VT Securitas class showcase (CLMS 370) to provide feedback on student presentations. 420 </div> 421 <div> 422 <em>Nov 2025: </em> Received the <strong>Top Technical Presentation Award</strong> at the <strong>CCI SWVA Graduate Student Summit</strong> for the talk <strong>âWhen Agents Test Code: Generative AI for Software Security.â</strong> 423 </div> 424 <div> 425 <em>Nov 2025: </em> Our paper <a href="https://arxiv.org/abs/2507.06092" target="_blank"><strong>âTaming Data Challenges in ML-based Security Tasksâ</strong></a> was accepted to <a href="https://asiaccs2026.cse.iitkgp.ac.in/home/" target="_blank"><strong>AsiaCCS 2026</strong></a>, one of only 74 papers accepted in cycle 1. 426 </div> 427 <div> 428 <em>July 2025: </em> 429 Check out our two new arXiv preprints: 430 <ul> 431 <li><a href="https://arxiv.org/abs/2507.06092" target="_blank">âTaming Data Challenges in ML-based Security Tasks: Lessons from Integrating Generative AIâ</a></li> 432 <li><a href="https://arxiv.org/abs/2507.05660" target="_blank">âTuneShield: Mitigating Toxicity in Conversational AI while Fine-tuning on Untrusted Dataâ</a></li> 433 </ul> 434 </div> 435 <div> 436 <em>Apr 2025: </em> I am now a <b>PhD candidate!</b> 437 </div> 438 <!-- NEW UPDATE END --> 439 <!-- <li> --> 440 <div><em>May 2024: </em> Pleasure to have attended <b>IEEE S&P 2024</b> as a coauthor of our paper on analysis of deepfake detection schemes.</div> 441 <!-- </li> --> 442 <!-- <li> --> 443 <div><em>Apr 2024: </em>Won and received the <b>IEEE S&P Student Travel Grant 2024</b>. </div> 444 <!-- </li> --> 445 <!-- <li> --> 446 <div><em>Mar 2024: </em>Presented a poster on <b>âFirst Look at Toxicity Injection Attacks on Open-domain Chatbots.â</b> at DMV Security Workshop 2024.</div> 447 <!-- </li> --> 448 <!-- <li> --> 449 <div><em>Dec 2023: </em> Pleasure to have attended <b>ACSAC 2023</b> as a coauthor of our paper on Data poisoning attacks in Dialogue based Learning (DBL) systems.</div> 450 <!-- </li> --> 451 <!-- <li> --> 452 <div><em>Oct 2023: </em>Presented on <b>âUsing GenAI to strengthen security defensesâ</b> at <b>VT Fall 2023 Skillshop Series</b> - Leveraging Creative Technologies.</div> 453 <!-- </li> --> 454 <!-- <li> --> 455 <div><em>Oct 2023: </em>Our research got featured in VPM News Focal Point - <a href="https://www.pbs.org/video/artificial-intelligence-what-are-the-risks-and-benefits-kaz2/"><b> âArtificial intelligence: What are the risks and benefits?â</a></b></div> 456 <!-- </li> --> 457 <!-- </ul> --> 458 </div> 459 </div> 460 </div> 461 </div> 462 </div> 463 <div id="research" class="container-fluid section-spacing"> 464 <h3 class="section-title">Research</h3> 465 <div class="panel panel-info" style="margin-right:20px"> 466 <div class="panel-body"> 467 <h4>Research Themes</h4> 468 <ul> 469 <li>Synthetic data pipelines for robust security ML under data scarcity or sensitivity.</li> 470 <li>LLM safety and toxicity mitigation for real-world conversational deployments.</li> 471 <li>Automated and adversarial security testing frameworks for evolving software and ML threats.</li> 472 <li>
472Evaluation pipelines emphasizing measurable security outcomes at deployment scale.</li> 473 </ul> 474 <h4>Representative Projects</h4> 475 <div class="project-block project"> 476 <h5 class="project-title"><a href="https://arxiv.org/abs/2507.06092" target="_blank">Improving Security ML Under Data Scarcity Using Generative AI</a> <span class="meta">(AsiaCCS 2026)</span></h5> 477 <p class="project-line"><span class="project-label">Problem:</span> Security ML classifiers are often limited by data scarcity and imbalance, rather than algorithmic design.</p> 478 <p class="project-line"><span class="project-label">Approach:</span> Applied <strong>Generative AI-based synthetic data augmentation</strong> to improve classifier generalization. Evaluated multiple state-of-the-art GenAI methods and introduced Nimai, a controlled synthetic data generation scheme for security tasks.</p> 479 <p class="project-line"><span class="project-label">Evaluation:</span> Conducted empirical evaluation across 7 diverse security tasks, including severely low-data settings.</p> 480 <p class="project-line"><span class="project-label">Outcome:</span> Achieved up to 32.6% performance improvement in security classifiers.</p> 481 </div> 482 <div class="project-block project"> 483 <h5 class="project-title"><a href="https://arxiv.org/abs/2507.05660" target="_blank">Reducing Toxicity in LLM Fine-tuning Pipelines</a> <span class="meta">(arXiv 2025)</span></h5> 484 <p class="project-line"><span class="project-label">Problem:</span> Fine-tuning LLMs on untrusted conversational data can introduce toxic or unsafe behavior.</p> 485 <p class="project-line"><span class="project-label">Approach:</span> Developed TuneShield, a defense framework for safe fine-tuning of <strong>LLaMA-based chat models</strong>, combining LLM-based toxicity detection, targeted synthetic data generation ("healing data"), and <strong>alignment using PEFT-based Direct Preference Optimization (DPO)</strong>.</p> 486 <p class="project-line"><span class="project-label">Evaluation:</span> Evaluated against toxicity injection, adaptive adversarial, and jailbreak attacks during conversational fine-tuning.</p> 487 <p class="project-line"><span class="project-label">Outcome:</span> Reduced toxic outputs to near-zero levels while preserving conversational quality.</p> 488 </div> 489 <div class="project-block project"> 490 <h5 class="project-title">Agent-based Automation for Security Test Generation <span class="meta">(Ongoing)</span></h5> 491 <p class="project-line"><span class="project-label">Problem:</span> Manual vulnerability testing for third-party libraries does not scale for software supply-chain security.</p> 492 <p class="project-line"><span class="project-label">Approach:</span> Built a <strong>self-correcting multi-agent LLM system</strong> for automated JUnit security test generation and execution.</p> 493 <p class="project-line"><span class="project-label">Evaluation:</span> Evaluated across 50 Java client codebases, using LLM-based validation to determine whether generated tests successfully trigger known vulnerabilities.</p> 494 <p class="project-line"><span class="project-label">Outcome:</span> Achieved automated end-to-end vulnerability test generation and validation using self-correcting LLM agents.</p> 495 </div> 496 </div> 497 </div> 498 </div> 499 <div id="section6" class="container-fluid section-spacing"> 500 <h3 class="section-title"><i class="fa-solid fa-building"></i> Experience</h3> 501 <div class="panel panel-info" style="margin-right:20px; text-align: justify;"> 502 <div class="panel-body"> 503 <!-- Section: Timeline --> 504 <section class="py-5"> 505 <!-- <ul> --> 506 <!-- <li class="timeline-item mb-5"> --> 507 <h4>Virginia Polytechnic Institute And State University</h4> 508 <p>Graduate Research Assistant</p> 509 <ul> 510 <li>Built evaluation-driven AI security systems spanning LLM safety and alignment, synthetic data generation, and adversarial ML, with publications at IEEE S&P, Asia CCS, and ACSAC.</li> 511 <li>Designed alignment-aware fine-tuning pipelines for LLaMA-based models using PEFT and Direct Preference Optimization (DPO) to mitigate toxicity injection and preserve conversational qual
511ity.</li> 512 <li>Developed multi-agent LLM systems that generate, execute, and self-correct JUnit security tests, enabling automated vulnerability validation across 50 Java client codebases in software supply-chain security.</li> 513 <li>Created controlled synthetic data generation frameworks that improved ML-based security classifiers by up to 32.6% under severe data scarcity.</li> 514 </ul> 515 516 <h4>JP Morgan Chase & IIIT Hyderabad</h4> 517 <p>Applied Research Collaboration</p> 518 <ul> 519 <li>Developed the first named-entity-labeled corpus for SEBI regulations, covering 7,500+ sub-regulations and defining 7 domain-specific entity types for financial compliance analysis.</li> 520 <li>Built an overlapping Named Entity Recognition system achieving 87.47% precision, enabling large-scale regulatory text understanding for governance and compliance workflows.</li> 521 </ul> 522 <!-- </li> --> 523 <!-- </ul> --> 524 </section> 525 <!-- Section: Timeline --> 526 </div> 527 </div> 528 </div> 529 530 <div id="section2" class="container-fluid section-spacing"> 531 <div id="publications"></div> 532 <h3 class="section-title"><i class="fa-solid fa-book"></i> Publications</h3> 533 <div class="panel panel-info" style="margin-right:20px"> 534 <div class="panel-body"> 535 <div class="row"> 536 <div class="col-md-12"> 537 <!-- <h5>Ph.D. Publications </h4> --> 538 <!-- <ul> --> 539 <!-- <li> --> 540 <div> 541 <strong>Optimus: A Robust Defense Framework for Mitigating Toxicity while Fine-Tuning Conversational AI</strong> 542 <br> Aravind Cheruvu, <b>Shravya Kanchi</b>, Sifat Muhammad Abdullah, Nicholas Kong, Daphne Yao, Murtuza Jadliwala, Bimal Viswanath <br>ACM CODASPY, Frankfurt am Main, Germany.<br> 543 <a href="https://arxiv.org/abs/2507.05660" target="_blank"> 544 <span class="label label-default">PDF</span> 545 </a> 546 </div> 547 <br> 548 <div> 549 <strong>Taming Data Challenges in ML-based Security Tasks: Lessons from Integrating Generative AI</strong> 550 <br><b>Shravya Kanchi</b>, Neal Mangaokar, Aravind Cheruvu, Sifat Muhammad Abdullah, Shirin Nilizadeh, Atul Prakash, Bimal Viswanath <br>ACM AsiaCCS 2026, Bangalore, India, 2026.<br> 551 <a href="https://arxiv.org/abs/2507.06092" target="_blank"> 552 <span class="label label-default">PDF</span> 553 </a> 554 <a href="https://github.com/secml-lab-vt/taming-data-challenges-security-ml/" target="_blank"> 555 <span class="label label-default">Code and dataset</span> 556 </a> 557 </div> 558 <br> 559 <!-- <li> --> 560 <div> 561 <strong>An Analysis of Recent Advances in Deepfake Image Detection in an Evolving Threat Landscape</strong> 562 <br> Sifat Muhammad Abdullah, Aravind Cheruvu, <b>Shravya Kanchi</b>, Taejoong Chung, Peng Gao, Murtuza Jadliwala and Bimal Viswanath <br>IEEE S&P (Oakland) 2024, San Francisco, CA, May 2024. <br> 563 <a href="https://arxiv.org/abs/2404.16212"> 564 <span class="label label-default">PDF</span> 565 </a> 566 <a href="https://github.com/secml-lab-vt/EvolvingThreat-DeepfakeImageDetect"> 567 <span class="label label-default">Code and dataset</span> 568 </a> 569 <a href="https://www.youtube.com/watch?v=Eg8Qb3zXdD4">
570 <span class="label label-default">Video</span> 571 </a> 572 </div> 573 <br> 574 <!-- </li> --> 575 <!-- <li> --> 576 <div> 577 <strong>A First Look at Toxicity Injection Attacks on Open-domain Chatbots</strong> 578 <br> Aravind Cheruvu, Connor Weeks, Sifat Muhammad Abdullah, <b>Shravya Kanchi</b>, Daphne Yao, and Bimal Viswanath <br>ACSAC 2023, Austin, Texas, December 2023. <br> 579 <a href="./Toxicity_injection_acsac_paper.pdf"> 580 <span class="label label-default">PDF</span> 581 </a> 582 <a href="https://github.com/secml-lab-vt/Chatbot-Toxicity-Injection/"> 583 <span class="label label-default">Code and dataset</span> 584 </a> 585 <a href="https://www.youtube.com/watch?v=Y9FTew96mxo"> 586 <span class="label label-default">Video</span> 587 </a> 588 </div> 589 <br> 590 <!-- </li> --> 591 592 <!-- </ul> 593 <h5>Prior Publications </h4> 594 <ul> --> 595 <!-- <li> --> 596 <div> 597 <strong>SEBI Regulation Biography</strong> 598 <br> Sathvik Sanjeev Buggana, Deepti Saravanan, <b>Shravya Kanchi</b>, Ujwal Narayan, Shivam Mangale, Lini T. Thomas, Kamalakar Karlapalem, Natraj Raman <br> WWW Workshop, Lyon, France, April 2022 <br> 599 <a href="./SEBI_regulation.pdf"> 600 <span class="label label-default">PDF</span> 601 </a> 602 </div> 603 <br> 604 <!-- </li> --> 605 <!-- <li> --> 606 <div> 607 <strong>A Multi Perspective Access Control in a Smart Home</strong> 608 <br><b>Shravya Kanchi</b>, Kamalakar Karlapalem <br> CODASPY, Online, April 2021 <br> 609 <a href="./codaspy_poster.pdf"> 610 <span class="label label-default">PDF</span> 611 </a> 612 <a href="https://github.com/shrave/SmartHomePrivilege-Manager"> 613 <span class="label label-default">Code</span> 614 </a> 615 </div> 616 <br> 617 <!-- </li> --> 618 <!-- </ul> --> 619 </div> 620 </div> 621 </div> 622 </div> 623 </div> 624 <div id="section5" class="container-fluid section-spacing"> 625 <h3 class="section-title"><i class="fa fa-cogs" aria-hidden="true"></i> 626 Skills</h3> 627 <div class="panel panel-info" style="margin-right:20px"> 628 <div class="panel-body"> 629 <div class="row"> 630 <div class="col-md-12"> 631 <p class="skills-title"><strong>AI Security & Generative AI</strong></p> 632 <p>AI Security, LLM Safety & Alignment, Generative AI, Adversarial Machine Learning, Security Evaluation & Benchmarking</p> 633 634 <p class="skills-title"><strong>Model Fine-Tuning & Alignment</strong></p> 635 <p>Direct Preference Optimization (DPO), PEFT (LoRA), Supervised Fine-Tuning (SFT), Instruction Tuning, Prompt Engineering</p> 636 637 <p class="skills-title"><strong>Synthetic Data & Robustness</strong></p> 638 <p>Synthetic Data Generation, Tabular Data Augmentation, Concept Drift, Robustness Analysis, Data Poisoning & Toxicity Injection</p> 639 640 <p class="skills-title"><strong>Agent-Based Systems</strong></p> 641 <p>LLM Agents, Multi-Agent Systems, Self-Correcting Agents, Automated Security Test Generation, Agentic Reasoning</p> 642 643 <p class="skills-title"><strong>Security Domains</strong></p> 644 <p>Malware & Phishing Detection, Malicious URL Detection, Network Intrusion Detection (IDS), Softw
644are Supply-Chain Security, Access Control Systems</p> 645 646 <p class="skills-title"><strong>ML Frameworks & Tooling</strong></p> 647 <p>PyTorch, Hugging Face Transformers, LangChain, Evaluation Pipelines</p> 648 </div> 649 </div> 650 </div> 651 </div> 652 </div> 653 654<div id="section3" class="container-fluid section-spacing"> 655 <h3 class="section-title"><i class="fa-solid fa-school"></i> Education</h3> 656 <div class="panel panel-info" style="margin-right:20px"> 657 <div class="panel-body"> 658 <!-- Section: Timeline --> 659 <section class="py-5"> 660 <ul class="timeline"> 661 <li class="timeline-item education-item"> 662 <h4>Virginia Polytechnic Institute And State University (Virginia Tech)</h4> 663 <div class="edu-row"> 664 <span class="edu-degree">Ph.D. in Computer Science</span> 665 <span class="edu-years">Aug 2021 - May 2026</span> 666 </div> 667 <div class="edu-cgpa">GPA: 3.9/4</div> 668 <!-- <p> 669 <b>Graduate courses:</b> Data Analytics, Deep Learning, Hot Topics in Security and AI, Theory of Algorithms, Advanced Machine Learning, Security risks of Generative AI, Defenses Against ML Powered Adversaries 670 </p> --> 671 </li> 672 <li class="timeline-item education-item"> 673 <h4>International Institute of Information Technology (IIIT), Hyderabad</h4> 674 <div class="edu-row"> 675 <span class="edu-degree">Masters by Thesis in Computer Science</span> 676 <span class="edu-years">Jul 2018 - May 2021</span> 677 </div> 678 <div class="edu-cgpa">GPA: 9/10</div> 679 <!-- <p>CGPA: 8.51/10.0 "Gold Medal for best outgoing student from the Department of I.T."</p> --> 680 </li> 681 <li class="timeline-item education-item"> 682 <h4>Indian Institute of Information Technology (IIIT), Sricity</h4> 683 <div class="edu-row"> 684 <span class="edu-degree">Bachelor of Technology (Honors) in Computer Science</span> 685 <span class="edu-years">Jul 2014 - May 2018</span> 686 </div> 687 <div class="edu-cgpa">GPA: 8.79/10</div> 688 <!-- <p>CGPA: 8.51/10.0 "Gold Medal for best outgoing student from the Department of I.T."</p> --> 689 </li> 690 </ul> 691 </section> 692 <!-- Section: Timeline --> 693 </div> 694 </div> 695 </div> 696 <div id="section4" class="container-fluid section-spacing"> 697 <h3 class="section-title"><i class="fa-solid fa-certificate"></i> Certifications</h3> 698 <div class="panel panel-info" style="margin-right:20px"> 699 <div class="panel-body"> 700 <div class="row"> 701 <div class="col-md-12"> 702 <div><strong><a href="https://ipxcourses.org/course-overview/" target="_blank">CopyrightX</a></strong> â US Copyright Law course jointly offered by Harvard Law School & <strong><a href="https://cyber.harvard.edu/" target="_blank">Berkman Klein Center for Internet and Society</a></strong>.</div> 703 </div> 704 </div> 705 </div> 706 </div> 707 </div> 708 709 <!-- <div id="section7" class="container-fluid"> 710 <h3 style="color: #ffffff;"><i class="fa-solid fa-video"></i> Videos</h3> 711 <div id="videoCarousel" class="carousel slide"> 712 <ol class="carousel-indicators"> 713 <li data-target="#videoCarousel" data-slide-to="0" class="active"></li> 714 <li data-target="#videoCarousel" data-slide-to="1"></li> 715 </ol> 716 <div class="carousel-inner"> 717 <div class="item active"> 718 <div class="embed-responsive"> 719 <iframe class="embed-responsive-item" src="https://www.youtube.com/embed/Y9FTew96mxo?si=YqeYCBQf6z4lAfwW" allowfullscreen></iframe> 720 </div> 721 </div> 722 <div class="item"> 723 <div class="embed-responsive"> 724 <iframe class="embed-responsive-item" src="https://player.pbs.org/viralplayer/3085363248/" allow="encrypted-media" allowfullscreen></iframe> 725 </div> 726 </div> 727 </div> 728 <a class="left carousel-control" href="#videoCarousel" role="button" data-sli
728de="prev"> 729 <span class="glyphicon glyphicon-chevron-left" aria-hidden="true"></span> 730 <span class="sr-only">Previous</span> 731 </a> 732 <a class="right carousel-control" href="#videoCarousel" role="button" data-slide="next"> 733 <span class="glyphicon glyphicon-chevron-right" aria-hidden="true"></span> 734 <span class="sr-only">Next</span> 735 </a> 736 </div> --> 737 </div> 738 </div> 739 </div> 740 </body> 741</html>
Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.